yolo4+visdrone2020数据集

和上一篇yolov3的流程差不多,只是更改yolo的cfg文件和预训练的权重不同，将操作移至./darknet/build/x64文件夹内，将./darknet/下的darknet执行文件复制到./darknet/build/x64/下

cfg主要更改

1. 分辨率为416*416，如果你的显卡内存不足可以调整subdivisions，减少每次放入内存图片的数量

2. max_batches=20000(class的2000倍，class=10)， steps= 16000,18000(max_batches的80%和90%)

3. [yolo] 部分，修改class=10，filters = 45 （(class+5)*3）

预训练权重可在官方github上下载

yolov4.cfg  -> yolov4.conv.137（预训练权重）

yolov4-tiny.cfg  -->  yolov4-tiny.conv.29

 

 

训练

./darknet detector train data/voc-custom.data cfg/yolov4-custom-visdrone.cfg yolov4.conv.137 2>&1 | tee visualization/train_yolov4.log     #yolov4.cfg,

./darknet detector train data/voc.data cfg/yolov4-tiny-visdrone.cfg yolov4-tiny.conv.29 2>&1 | tee visualization/train_yolov4.log　　　　　　#yolov4-tiny.cfg,

 暂停后继续训练(将backup文件夹下的yolov4-***_last.weights复制到x64文件夹下)

./darknet detector train data/voc.data cfg/yolov4-tiny-visdrone.cfg yolov4-tiny-visdrone_last.weights 2>&1 | tee visualization/train_yolov4.log

./darknet detector train data/voc.data cfg/yolov4-visdrone.cfg yolov4-custom-visdrone_last.weights 2>&1 | tee visualization/train_yolov4.log

 

测试视频

./darknet detector demo data/voc-custom.data cfg/yolov4-custom-visdrone.cfg yolov4-custom-visdrone_last.weights -ext_output cardemo.mp4