video-classification-3d-cnn-pytorch
video-classification-3d-cnn-pytorch copied to clipboard
kenshohara
Classification on UCF101
I'm trying to classify based on my own setup for UCF-101, where I added one of my own classes and trained using the 3D-Resnets-PyTorch library.
This is the command I run:
python main.py --input ./input --video_root ./input_vids --output ./output.json --model ../3D-ResNets-PyTorch/data/UCF-Results/save_280.pth --mode score
I'm getting this error:
KeyError: 'unexpected key "module.layer2.0.downsample.0.weight" in state_dict'
I found something here which says that nn.DataParallel is the issue when loading the module, but I can't get the nn.DataParallel or the suggested solution working: https://stackoverflow.com/questions/44230907/keyerror-unexpected-key-module-encoder-embedding-weight-in-state-dict
I'm also wondering if maybe it's that there is no downsample in this library's ResNet? I'm using ResNet34, so the previously closed Github Issue isn't the same.
Has this ever been run on one of the other datasets successfully? I appreciate any feedback!
Which model do you use? Do you use ResNet-34 with shortcut type A? If you use ResNet-34 with shortcut type B, please specify --resnet_shortcut B.
