Dataset · google-research-datasets
Web images with alt-text captions for image-to-text
google-research-datasets/conceptual_captions
~3.3M web images with alt-text captions for image-to-text training.
What's inside
About 3.3M images with text descriptions in English. Captions and images were collected from the web, with original descriptions sourced from the HTML alt attribute. Styles are more diverse than in curated caption sets.
What you can do with it
- Training an image captioning model
- Fine-tuning a CLIP-style image-text model
- Evaluating description quality on web images
- Tasks
- Image → text
- Languages
- English
- Size
- 1M–10M rows
- Downloads this month
- 11.6K
- Likes
- 112
Models trained on it
- hanxunh/clip_backdoor_vit_b16_cc3m_badnetsZero shot image classification 52
- hanxunh/clip_backdoor_rn50_cc3m_clean_labelZero shot image classification 43
- hanxunh/clip_backdoor_rn50_cc3m_badnetsZero shot image classification 39
- hanxunh/clip_backdoor_vit_b16_cc3m_sigZero shot image classification 38
- hanxunh/clip_backdoor_rn50_cc12m_clean_labelZero shot image classification 30
- hanxunh/clip_backdoor_rn50_cc3m_wanetZero shot image classification 29
- hanxunh/clip_backdoor_vit_b16_cc3m_nashvilleZero shot image classification 26
- hanxunh/clip_backdoor_rn50_cc3m_blendZero shot image classification 25
How to load it
from datasets import load_dataset
ds = load_dataset("google-research-datasets/conceptual_captions")