In the field of underwater image recognition, a chip with smaller footprint and lower energy consumption is required to be implanted into autonomous intelligent underwater vehicle to make real-time response to the surrounding objects. Therefore, a promising accelerator with high performance and low energy consumption is designed, which optimizes the features possessed by convolutional neural network. The sharing of weights between neurons reduces the memory requirement. With all convolutional neural network data stored within on-chip static random-access memory, the need for memory access is drastically decreased. Besides, several small processing elements are used to form neural functional unit, which considerably reduces the bandwidth requirement through inter-processing element data transmission. By sending control signals to autonomous underwater vehicle, this accelerator enables it to avoid dangerous areas such as rocks and algae in time. The result suggests the proposed accelerator successfully achieves a higher processing speed than that of CPU and GPU with a footprint of 6.09 mm2 only and the energy consumption of 327.3 mW at 1 GHz.
WuZKarimiHRDangC. An approximation algorithm for graph partitioning via deterministic annealing neural network. Neural Netw2019; 117: 191–200.
2.
WuZKarimiHRDangC. A Deterministic Annealing Neural Network Algorithm for the Minimum Concave Cost Transportation Problem. IEEE Trans Neural Netw Learn Syst. Epub ahead of print 19December2019. DOI: 10.1109/TNNLS.2019.2955137.
3.
MarshLCopleyJTHuvenneVA, et al. Getting the bigger picture: using precision Remotely Operated Vehicle (ROV) videography to acquire high−definition mosaic images of newly discovered hydrothermal vents in the Southern Ocean. Deep Sea Res Part II Top Stud Oceanogr2013; 92: 124–135.
4.
JiangYZhaoMHuC, et al. A parallel FP-growth algorithm on World Ocean Atlas data with multi-core CPU. J Supercomput2019; 75(2): 732–745.
5.
ZerrBMailfertGBertholomA, et al. Sidescan sonar image processing for AUV navigation. In: Europe Oceans 2005, Brest, 20–23June2005, vol. 1, pp.124–130. New York: IEEE.
6.
SoaresJMAguiarAPPascoalAM, et al. Triangular formation control using range measurements: an application to marine robotic vehicles. In: 3rd IFAC workshop on navigation, guidance and control of underwater vehicles (NGCUV 2012), Porto, 10–12 April 2012, pp.1–6. New York: IFAC.
7.
WangCJiangYWangK, et al. A field-programmable gate array system for sonar image recognition based on convolutional neural network. Proc IMechE Part I: J Systems and Control Engineering. Epub ahead of print 15 July 2020. DOI: 10.1177/0959651820939345.
8.
LeCunYBengioYHintonG. Deep learning. Nature2015; 521(7553): 436–444.
9.
BilenHFernandoBGavvesE, et al. Dynamic image networks for action recognition. In: Proceedings of the IEEE conference on computer vision and pattern recognition 2016, Las Vegas, NV, 27–30 June 2016, pp.3034−3042. New York: IEEE.
10.
CiresanDCMeierUMasciJ, et al. Flexible, high performance convolutional neural networks for image classification. In: Twenty-second international joint conference on artificial intelligence, Barcelona, 16–22 July 2011, pp.1237–1242. New York: ACM.
11.
VanhouckeVSeniorAMaoMZ. Improving the speed of neural networks on CPUs. In: Deep learning and unsupervised feature learning workshop (NIPS 2011), Granada, 16–17 December 2011, pp.1–8. San Diego, CA: NeurIPS Foundation.
12.
DegalahalVTuanT. Methodology for high level estimation of FPGA power consumption. In: Proceedings of the 2005 Asia and South Pacific design automation conference, Shanghai, China, 18–21 January 2005, pp.657–660. New York: IEEE.
13.
HameedRQadeerWWachsM, et al. Understanding sources of inefficiency in general-purpose chips. In: Proceedings of the 37th annual international symposium on Computer architecture, Saint-Malo, 19−23 June 2010, pp.37−47. New York: ACM.
14.
ZhaoMHuCWeiF, et al. Real-time underwater image recognition with FPGA embedded system for convolutional neural network. Sensors2019; 19(2): 350.
15.
NaneRSimaVMPilatoC, et al. A survey and evaluation of FPGA high-level synthesis tools. IEEE Trans Comput Aided Design Integr Circuits Syst2015; 35(10): 1591–1604.
16.
FarabetCLeCunYKavukcuogluK, et al. Large-scale FPGA-based convolutional networks. In: BekkermanRBilenkoMLangfordJ (eds) Scaling up machine learning: parallel and distributed approaches. Cambridge: Cambridge University Press, 2011, pp.399−419.
17.
ChoiYKCongJFangZ, et al. A quantitative analysis on microarchitectures of modern CPU-FPGA platforms. In: Proceedings of the 53rd annual design automation conference, Austin, TX, 5−9 June 2016, article no. 109, pp.1−6. New York: IEEE.
18.
ChenTChenYDurantonM, et al. BenchNN: on the broad potential application scope of hardware neural network accelerators. In: 2012 IEEE international symposium on workload characterization (IISWC), La Jolla, CA, 4−6 November 2012, pp.36−45. New York: IEEE.
19.
BelhadjBJoubertALiZ, et al. Continuous real-world inputs can open up alternative accelerator designs. In: Proceedings of the 40th annual international symposium on computer architecture, Tel-Aviv, Israel, 23−27 June 2013, pp.1−12. New York: ACM.
20.
LeeSYAggarwalJK. Parallel 2-D convolution on a mesh connected array processor. IEEE Trans Pattern Anal Mach Intell1987; 9: 590−594.
21.
KhanMMLesterDRPlanaLA, et al. SpiNNaker: mapping neural networks onto a massively-parallel chip multiprocessor. In: 2008 IEEE international joint conference on neural networks (IEEE world congress on computational intelligence), Hong Kong, China, 1−8 June 2008, pp.2849−2856. New York: IEEE.
22.
JarrettKKavukcuogluKRanzatoMA, et al. What is the best multi-stage architecture for object recognition? In: 2009 IEEE 12th international conference on computer vision, Kyoto, Japan, 29September−2October2009, pp.2146−2153. New York: IEEE.
23.
KrizhevskyASutskeverIHintonGE. Imagenet classification with deep convolutional neural networks. In: Advances in neural information processing systems 2012, Lake Tahoe, NV, 3−8December2012, pp.1097−1105. New York: ACM.
24.
SchmidhuberJ. Deep learning in neural networks: an overview. Neural Netw2015; 61: 85–117.
25.
FarabetCMartiniBCordaB, et al. Neuflow: a runtime reconfigurable dataflow processor for vision. In: CVPR 2011 workshops, Colorado Springs, CO, 20−25 June 2011, pp.109−116. New York: IEEE.
26.
DuZPalemKLingamneniA, et al. Leveraging the error resilience of machine-learning applications for designing highly energy efficient accelerators. In: 2014 19th Asia and South Pacific design automation conference (ASP-DAC), Singapore, 20−23 January 2014, pp.201−206. New York: IEEE.
27.
TemamO. A defect-tolerant accelerator for emerging high-performance applications. In: 2012 39th Annual international symposium on computer architecture (ISCA), Portland, OR, 9−13 June 2012, pp.356−367. New York: IEEE.
28.
HoliJLHwangJN. Finite precision error analysis of neural network hardware implementations. IEEE Trans Comput1993; 42(3): 281–290.
29.
LarkinDKinaneAO’ConnorN. Towards hardware acceleration of neuroevolution for multimedia processing applications on mobile devices. In: ICONIP 2006, Hong Kong, China, 3−6 October 2006, pp.1178−1188. Berlin: Springer.
30.
ChakradharSSankaradasMJakkulaV, et al. A dynamically configurable coprocessor for convolutional neural networks. In: Proceedings of the 37th annual international symposium on computer architecture, Saint-Malo, 19−23 June 2010, pp.247−257. New York: ACM.
31.
KingEJSwartzlanderEE. Data-dependent truncation scheme for parallel multipliers. In: Conference record of the thirty-first Asilomar conference on signals, systems and computers (Cat. No. 97CB36136), Pacific Grove, CA, 2−5 November 1997, vol. 2, pp.1178−1182. New York: IEEE.
