Dataset · pixparse
12 million image-text description pairs
pixparse/cc12m-wds
Large Conceptual Captions 12M dataset in WebDataset format for pre-training multimodal models.
What's inside
The dataset contains 12 million image and text caption pairs. It is an expanded version of the Conceptual Captions 3M (CC3M) corpus, collected using softer filtering rules. The data is packaged into WebDataset (.tar) shards for streaming reading during training.
What you can do with it
- Pre-training vision-language multimodal models (CLIP, VLM)
- Training image-to-text generation models
- Tasks
- Image → text
- Size
- 10M–100M rows
- Downloads this month
- 38.2K
- Likes
- 45
Models trained on it
- KBlueLeaf/TIPO-v2.1-1B-A200MText generation 25.1K
- KBlueLeaf/TIPOv2-1B-A200MText generation 1.4K
- AuroraAI-Research/AuroraIMG-6MText → image 86
- rwightman/naflex_ViT-B-32.cc12mZero shot image classification 47
- rwightman/naflex_ViT-B-32.cc12m-pj24-32-40Zero shot image classification 38
- rwightman/mammut2-moderntext_ViT-B-32.cc12m-qk-lr3e3Zero shot image classification 35
- rwightman/mammut2_ViT-B-32.cc12mZero shot image classification 28
- rwightman/mammut2-naflex_ViT-B-32.cc12mZero shot image classification 28
How to load it
from datasets import load_dataset
ds = load_dataset("pixparse/cc12m-wds")