1 collection of PyTorch-compatible video datasets, transforms, and samplers designed for deep learning workflows. The library provides standardized interfaces for loading video files and applying synchronized spatial and temporal augmentations across frame sequences.
Use Cases
- Train action recognition models using the integrated video dataset classes and temporal samplers
- Apply synchronized spatial transforms across multiple video frames using the library's augmentation modules
- Develop custom video processing pipelines by leveraging the provided sampler and transform interfaces
Strengths
- Includes specialized temporal samplers for frame selection and sequence sampling
- Provides video-specific transforms for synchronized spatial and temporal data augmentation
- Implements standardized PyTorch Dataset classes for efficient video file handling