Long multi-view videos collected for the SIGGRAPH Asia 2024 (TOG) paper: Representing Long Volumetric Video with Temporal Gaussian Hierarchy.
Camera parameter convensions follow EasyVolcap. Some sequences contain an extra synchronization correction list (time computed from frame index - sync.json = actual timestamp).
We also provide a set of point clouds extracted from multiview images using various tools like COLMAP and RealityCapture… See the full description on the dataset page:
https://huggingface.co/datasets/zju3dv/SelfCap-Dataset.
