|
--- |
|
license: cc-by-4.0 |
|
--- |
|
Video Test-Time Adaptation for Action Recognition (CVPR 2023) |
|
|
|
- [Project Page](https://wlin-at.github.io/vitta) |
|
- [GitHub Repo](https://github.com/wlin-at/ViTTA) |
|
- [Arxiv Paper](https://arxiv.org/abs/2211.15393) |
|
|
|
### Dataset Description |
|
This dataset repo contains the following two datasets: |
|
- `Kinetics400_val_corruptions`: 12 corruption types for the 19877 validation videos on Kinetics400. |
|
- `SSv2_val_corruptions`: 12 corruption types for the 24777 validation videos on Something-Something v2. |
