Fast Classification and Detection of Marine Targets in Complex Scenes with YOLOv3

In order to meet the needs of fast detection and classification of different marine targets during intelligent unmanned surface vehicle (USV) operations, In this paper, I introduce a convolutional neural network based on one of the most effective object detection algorithms, named YOLOv3, to classify and detect images of different marine targets. Firstly, I showed the network structure of the algorithm in this paper. Then, I explained how I got the optimal anchor box parameter of the algorithm. Finally, I improved the activation function to make the algorithm more robust to noise. The final results show that the MAP of the detector in this paper is 91.83%,and we reach a detection rate of 58.3 fps by improving the YOLOV3 algorithm.

[1]  Ali Farhadi,et al.  You Only Look Once: Unified, Real-Time Object Detection , 2015, 2016 IEEE Conference on Computer Vision and Pattern Recognition (CVPR).

[2]  Kaiming He,et al.  Faster R-CNN: Towards Real-Time Object Detection with Region Proposal Networks , 2015, IEEE Transactions on Pattern Analysis and Machine Intelligence.

[3]  Ali Farhadi,et al.  YOLO9000: Better, Faster, Stronger , 2016, 2017 IEEE Conference on Computer Vision and Pattern Recognition (CVPR).

[4]  Trevor Darrell,et al.  Rich Feature Hierarchies for Accurate Object Detection and Semantic Segmentation , 2013, 2014 IEEE Conference on Computer Vision and Pattern Recognition.

[5]  Ali Farhadi,et al.  YOLOv3: An Incremental Improvement , 2018, ArXiv.

[6]  Sepp Hochreiter,et al.  Fast and Accurate Deep Network Learning by Exponential Linear Units (ELUs) , 2015, ICLR.

[7]  Xiu Li,et al.  Deep residual networks for plankton classification , 2016, OCEANS 2016 MTS/IEEE Monterey.

[8]  Xiu Li,et al.  Accelerating fish detection and recognition by sharing CNNs with objectness learning , 2016, OCEANS 2016 - Shanghai.