MobileFaceNet_Pytorch
MobileFaceNet_Pytorch copied to clipboard
MobileFaceNets: Efficient CNNs for Accurate Real-Time Face Verification on Mobile Devices
Hi @Xiaoccer , Since I am not able to download the dataset (I have no pan.baidu account), I want to know how you created the CASIA-WebFace-112x96.txt . What is the...
Hi author, How can we set the mean and threshold for 2 input faces? For instance, we firstly calculate the 2 features of the input faces. Then we have to...
测试精度
请问为什么我自己训练得不到那么高的精度?测试精度只有60%左右
在训练时loss一直在下降,突然变成nan了,请问一下有没有什么解决的方法呢?
博主给的代码保存断点似乎没有效果,和重新训练是一模一样的,为什么会这样呢?
模型提取的128维人脸特征向量数值有正负,并且有大于1的。数值不是0-1之间吗?求解。 谢谢!
你好!请问一下为什么从头训练以后,batch取48,训练35个epoch,loss缓慢下降,平均准确率一直显示是50%没变化过,请问一下是什么问题呢?
It is possible to download only if you have Chineese phone number. Can you provide other cloud? Also I can create a folder on my Google Drive and you can...