The MagicData Dataset, introduced in MagicMotion, is a comprehensive trajectory controllable video generation dataset with rich semantic annotations. This dataset features high-quality videos with precise segmentation masks and bounding boxes annotations, designed to advance the field of controllable video synthesis and understanding.
The dataset consists of 23K diverse videos, each meticulously… See the full description on the dataset page: https://huggingface.co/datasets/quanhaol/MagicData.
