ATD-12K is a large-scale animation triplet dataset, which comprises 12,000 triplets(train10k,test2k) by manually inspect and the test2k with rich annotations, including levels of difficulty, the Regions of Interest (RoIs) on movements, and tags on motion categories
The dataset collected from 30 series of movies(Is wild and modern) made by diversified producers, with a total duration of 25+ hours. total of 101 clips in two resolutions (i.e., 1920×1080, 1280×720)
Variants: ATD-12K
This dataset is used in 1 benchmark:
| Task | Model | Paper | Date |
|---|---|---|---|
| Video Frame Interpolation | M2M-PWC | Many-to-many Splatting for Efficient Video … | 2022-04-07 |
Recent papers with results on this dataset: