摘 要：文章基于 YOLOv4 进行了蛇类检测。在谷歌的 Open Image 数据集中下载已标注蛇类图片，使用谷歌的 Colab 平台进行实验，在 Darknet 框架下对网络模型进行训练。经对比，YOLOv4 的最终性能高于常见的 onestage 检测算法，相近准确度下速度快于 twostage 检测算法。最终结果显示，YOLOv4 在识别蛇类图像时准确度达 95.55%，平均检测时间为 37 ms，帧处理速率达 27FPS（帧 / 秒）。该检测速度和检测精度满足大部分背景下蛇类检测的需求，使蛇类检测与识别具备了可行性。
关键词：深度学习；卷积神经网络；蛇类；YOLOv4；目标检测
DOI:10.19850/j.cnki.2096-4706.2021.13.009
中图分类号：TP391.4 文献标识码：A 文章编号：2096-4706（2021）13-0034-03
Snake Image Recognition Based on YOLOv4
WANG Boxin, LI Dan
(School of Computer and Software, Jincheng College of Sichuan University, Chengdu 611731, China)
Abstract: In this paper, snake detection is carried out based on YOLOv4. Download the marked snake pictures in Google, s Open Image dataset, conduct experiments using Google, s Colab platform, and train the network model under the Darknet framework. After comparison, the final performance of YOLOv4 is higher than the common onestage detection algorithm, and the speed of it is faster than the twotage detection algorithm in the context of similar accuracy. The final results show that the accuracy of YOLOv4 in identifying snake images reaches 95.55%, the average detection time is 37 ms, and the frame processing rate reaches 27 FPS (frames/ second).The detection speed and accuracy meet the needs of snake detection in most backgrounds, which makes snake detection and recognition feasible.
Keywords: deep learning; convolutional neural network; snake; YOLOv4; object detection
参考文献：
[1] GIRSHICK R，DONAHUE J，DARRELL T，et al. Rich Feature Hierarchies for Accurate Object Detection and Semantic Segmentation [C]//2014 IEEE Conference on Computer Vision and Pattern Recognition.Columbus：IEEE，2014：580-587.
[2] HE K，ZHANG X Y，REN S Q ，et al. Spatial Pyramid Pooling in Deep Convolutional Networks for Visual Recognition [J]. IEEE Transactions on Pattern Analysis and Machine Intelligence， 2015，37（9）：1904-1916.
[3] MEI Y Q，FAN Y C，ZHANG Y L，et al. Pyramid Attention Networks for Image Restoration [EB/OL].[2020-04-28].https://www. researchgate.net/publication/341040607.
[4] BOCHKOVSKIY A，WANG C Y，LIAO H Y M． YOLOv4：Optimal Speed and Accuracy of Object Detection [J/ OL].（2020-04-23）.arXiv：2004.10934 [cs.CV].https://arxiv.org/ abs/2004.10934.
[5] ZHENG Z，WANG P，LIU W，et.al. DongweiRen.DistanceIoULoss：Faster and Better Learning for Bounding Box Regression [C]// Proceedings of the AAAI Conference on Artificial Intelligence.2020，34 （7）：12993-13000.
[6] REDMON J，FARHADI A. YOLOv3： An Incremental Improvement [J/OL].（2018-04-08）.arXiv：1804.02767 [cs.CV]. https://arxiv.org/abs/1804.02767.
[7] REN S，HE K，GIRSHICK R，et al. Faster R-CNN： Towards Real-Time Object Detection with Region Proposal Networks [J].IEEE Transactions on Pattern Analysis and Machine Intelligence， 2017，39（6）：1137-1149.
作者简介：王博鑫（2000.04—），男，汉族，四川西昌人， 本科在读，研究方向：软件工程。