FFC
FFC copied to clipboard
Official code for fast face classification
can i get the pretrained weight ?
May I ask where the data enhancement method mentioned in README.md is reflected? Could you help me to guide Xiao Bai thank you
您好,感谢您的工作。有些疑问希望可以得到您的解答: 1. 原文中阐述了instance-based的两个问题,但是第二个问题是否已经被CVPR 2021 DCQ那篇文章解决?第一个问题在DCQ中仍然存在,但是改进了这个问题好像结果并没有比DCQ好? 2. 原文中阐述的这两个问题,是如何影响实验结果呢?更具体来说,您定义的更新速度这个变量会影响最终识别的结果吗,更新快意味着结果好吗?【1. If the number of identities is severely imbalanced, the update speed of the identities’ centers that have rare number of images is too...
Indeed, I wish FFC can solve my problem of GPU memory limitation without damaging the model performance. so can you describe more details about the difference/advantage of FFC comparing with...
if world_size=8 then error: argument world_size: invalid int value for one server which have 8 gpu cart, the value of "world_size" can be setted as what? thanks
Dear @tiandunx Thank you for your work. In the Virtual FC paper at CVPR 2021, we have to change the network HEAD and replace it with the Virtual FC layer...
in Main.py line 110 : optim_config = load_config('config/optim_config') there is no load_config defined