VidaForge/VidaForge-3M

O dataset VidaForge/VidaForge-3M está entre os destaques do Hugging Face — dados que alimentam o treinamento e a avaliação dos modelos do momento.

Ficha do dataset

Sobre o dataset

3.14 million scene-level video clips with multi-level captions, camera labels, semantic tags, quality signals, and duplicate groups. VidaForge Code · Project Blog · Source Dataset Overview VidaForge-3M is a large-scale video pretraining dataset produced with VidaForge, an open data pipeline for building and studying video foundation model pretraining data. The release contains 3,141,246 annotation-complete clips totaling 6,475.1 hours. Every clip has four… See the full description on the dataset page: https://huggingface.co/datasets/VidaForge/VidaForge-3M.

Como carregar

Use a biblioteca datasets do Hugging Face:

pip install -U datasets

Como é um dataset grande, vale carregar em modo streaming (sem baixar tudo):

from datasets import load_dataset

ds = load_dataset("VidaForge/VidaForge-3M", split="train", streaming=True)
for exemplo in ds.take(3):
    print(exemplo)

Tags

text-to-video video-text-to-text video-classification video video-pretraining video-captioning video-generation self-supervised-learning

Explorar o dataset no Hugging Face →