Sign in to view source links and access this dataset
Description
Syn4D is a large-scale, fully-synthetic multiview dataset of dynamic scenes. It provides dense, complete, and accurate geometric annotations, including per-pixel depth maps, multi-view camera trajectories, dense long-range 3D point tracks, and parametric SMPL-X human body annotations. The dataset was created by harpreetsahota and was last updated on 2026-05-26.
Use Cases
4D reconstruction of dynamic scenes based on multi-view camera trajectories and dense point tracks
Depth estimation and 3D point tracking based on per-pixel depth maps and long-range 3D point tracks
Novel-view synthesis for dynamic scenes based on multi-view RGBD data
Human pose estimation based on parametric SMPL-X body annotations
Strengths
Dataset is described as large-scale and fully-synthetic
Provides dense, complete, and accurate geometric annotations
Includes multi-view camera trajectories and dense long-range 3D point tracks
Limitations
Column-level documentation is absent; field semantics must be inferred after download
Row count is unknown, which may limit suitability assessment
Data may reflect bias inherent to synthetic generation methods
Provenance
Source
huggingface
Collection Method
Fully-synthetic generation
Freshness
Last updated 2026-05-26 22:23:07; freshness should be verified
License is unknown; users should verify terms of use before downloading.