zengwb-lx / face-tracking-usingfairmot Goto Github PK

你好！我用你提供的代码进行训练时，第一个epoch收敛是正常的，loss逐渐下降到4.3左右，但到了第二个epoch loss突然变为nan

hi， great work , can it run on window10? ,thanks

Hi, many thanks for your great contribution! I wanna ask what makes the output model size change? Your uploaded pretrained model is 29 MB, but when I train on other datasets, the size changed to 40MB.

Some factors may make a difference:
input image size, object category, whether use multi-scale input training and fpn?
Which will influence the output model size?

转caffe 或onnx 推理时hm 这一层解析出来的值经过sigmoid后都很小（都小于0.3），请问您这边部署是转的什么模型？

Hi, thank you for your open code! I am using your code, about some details that i hope to communicate with you further. Looking forward to your reply. My weixin: 18374967041

First, thanks your finish an amazing project.My problems describe as title.Is there something I didn't do that made it slow? Or that's the velocity of this model.
Looking forward to your reply.

请想问下有没有casia-webface的label可以提供的，万分感谢

您好！在测试model_41epoch.pth的时候报错了，RuntimeError: ../models/model_41epoch.pth is a zip archive (did you mean to use torch.jit.load()?)

您好，可以提供一下转换成的onnx模型吗？
我使用torch.onnx.export导出onnx模型一切正常，再使用inferenceSession推理，精度非常非常低。与pth结果相差很远。
可以告知是为什么吗？

I have used webface dataset , but the id loss hard to convergence..
Have some tricks?