Skip to content

Dataset · VeroLM

Short Video Clips with Active Object Motion

VeroLM/VeroMotion

A filtered selection of 10,000 to 100,000 dynamic video clips without static scenes for training world models and video generation.

What's inside

The dataset contains 10K to 100K short single-shot video clips supplemented with text and tabular metadata. Static frames like talking heads or slide shows are intentionally excluded, and camera motion is filtered during evaluation, leaving only clips with continuous motion of the objects themselves.

What you can do with it

  • Training and testing world models for next-frame prediction
  • Training video generation models (text-to-video / image-to-video) for more realistic motion physics
  • Benchmarking tracking algorithms and object dynamics analysis in video streams
Size
10K–100K rows
Downloads this month
38.9K
Likes
1
Access
on request on Hugging Face

How to load it

from datasets import load_dataset

ds = load_dataset("VeroLM/VeroMotion")