Skip to content

Dataset · 3DTopia

Large-scale video dataset with pseudo-4D annotations

3DTopia/4DNeX-10M

Monocular videos of static and dynamic scenes with 4D annotations for 3D and 4D generation.

What's inside

A large hybrid dataset from the 4DNeX research paper, combining monocular videos from various sources featuring static and dynamic scenes. Videos are augmented with pseudo-4D annotations obtained using modern 3D and 4D reconstruction methods. The metadata specifies the image-to-3d task and a scale of over 1 trillion units.

What you can do with it

  • Train models to generate dynamic 4D scenes from images or videos
  • Joint modeling of RGB video and spatiotemporal geometry
  • Research in 3D reconstruction of moving objects
Tasks
Image → 3d
Size
over 1T rows
Downloads this month
31.5K
Likes
13

How to load it

from datasets import load_dataset

ds = load_dataset("3DTopia/4DNeX-10M")