CCSS1903 comments

Results 15 comments of


                                            CCSS1903

size mismatch for fc8.bias: copying a param with shape torch.Size([487]) from checkpoint, the shape in current model is torch.Size([101])

pretrained model应该不是用得ucf101

Train from scratch

有人解决了吗，训练精度一直很低

Train from scratch

![image](https://user-images.githubusercontent.com/56211787/96326764-10d47600-1066-11eb-9116-7b00ec5c74a5.png)

VideoDataset has a bug about data normalization

How should the code be modified?Training loss is always NAN

Y'and Vy'

You said FrameC is Xb， but the inputs to the net are FrameA and FrameB，isXa and Xb‘. FrameB are perturded frames.So Amplified Frames in dataset means Y’ not Y^？

Why should do this:frame[:] = roi_frame * (1. / 255)

归一化

Can you upload a pre-trained model?

@eugenelet pretrained model

issues of test

第一个代码出现了全黑的情况

拉取rtsp流解码报错

遇到了同样的问题，拉取的是rtsp流

dataset

Can ICA,POS CHROM and other traditional method codes be provided？