search

ChenYingpeng / darknet2caffe

Convert darknet weights to caffemodel
183 stars 88 forks source link

运行环境问题 #24

Open Larry-Zheng opened 3 years ago

Larry-Zheng commented 3 years ago

你好 我想问一下 这里存不存在大小端的问题 我是在windows下跑的代码 模型效果略微有些出入 我查了一下说weights文件是小端 不知道是不是这个问题

ChenYingpeng commented 3 years ago

Maybe. I have tested yolov4 in my other repo, which's results indicate that there is deviation. You can retrain about 1000 iters in caffe to restore accuary. Good luck.

Larry-Zheng commented 3 years ago

Maybe. I have tested yolov4 in my other repo, which's results indicate that there is deviation. You can retrain about 1000 iters in caffe to restore accuary. Good luck.

ok，thank you. By the way, we've been working on the yolo-fastest algorithm applying on huawei devices. We converted the model into caffe. The performance was passablely good. We really apprentice you and your work.