Project Structure

The repository is organized as follows:

• WLASL100_videos: Folder containing video data used for training and evaluation.
• nslt_100.json: Annotation file mapping videos to their corresponding labels and subsets (train, validation, test).
• TimeSformer.ipynb: Notebook implementing the TimeSformer model for ASL recognition.
• VideoMAE.ipynb: Initial implementation of VideoMAE for ASL recognition.
• VideoMAE_correct_split.ipynb: Corrected implementation of VideoMAE with proper dataset splits.
• VideoMAE_correct_split.pth: Pretrained weights for the corrected VideoMAE implementation.
• .gitattributes: Configuration for handling large files in GitHub.

Other trained models and configurations are available upon request.