action-recognition-using-3d-resnet

Use 3D ResNet to extract features of UCF101 and HMDB51 and then classify them.

how to use

1. Clone this repo:

   git clone https://github.com/vra/action-recognition-using-3d-resnet.git

2. Download 3D ResNet

3. Download its pretrained models, put these models to this repo's data/models/

4. run the script under scripts under to extract 3D resnet features of UCF101 and HMDB51:

   bash scripts/extract_resnet_3d_features.sh /path/to/video-classification/3d-cnn-pytorch ucf101 /path/to/ucf101/videos 
   bash scripts/extract_resnet_3d_features.sh /path/to/video-classification/3d-cnn-pytorch hmdb51 /path/to/hmdb51/videos 

Also, you can download my extracted features of ucf101 and hmdb51 at here and here. Remember to put the first one to data/jsons/ucf101 before you download the second one, otherwise the first one will be convered.

5. Run main.py to classify extracted 3D resnet features:

   python main.py -dataset hmdb51

Results:

strategy	dataset	accuracy
mean	ucf101	0.8487
max	ucf101	0.8667
mean	hmdb51	0.5425
max	hmdb51	0.5399