yakhyo / video-classification-pytorch

Video Classification using R(2+1)D based on ResNet18 on UCF-101 dataset. PyTorch Implementation.

Video Classification using R2Plus1D PyTorch model on UCF-101 dataset

Dataset:

• Download the dataset from here
• Prepare the folders as shown below:

├── UCF-101 
    ├── annotations
         ├── classInd.txt
         ├── testlist01.txt
         ├── trainlist01.txt
             ...
          
    ├── videos
         ├── ApplyEyeMakeup
         ├── ApplyLipstick
         ├── Archery
             ...

Train

bash train.sh

train.sh:

python3 -m torch.distributed.run --nproc_per_node=2 train.py --amp --cache-dataset

Test

bash test.sh

test.sh:

python3 test.py --cache-dataset