Dataset · Lin-Chen
1.2M image captions by GPT-4V
Lin-Chen/ShareGPT4V
Multimodal English captions for image-text alignment and training captioning models.
What's inside
ShareGPT4V Captions 1.2M is a set of multimodal captions generated using GPT-4 Vision. The data is in English, sized from 1 to 10 million examples, and modality in metadata is listed as text. The card also mentions the related ShareGPT4Video with 40K and 4.8M captions.
What you can do with it
- Train a vision-language model for caption generation
- Fine-tune LoRA for image captioning
- Evaluate modality alignment
- Tasks
- Visual question answering, Question answering
- Languages
- English
- Size
- 1M–10M rows
- Downloads this month
- 2.2K
- Likes
- 328
Models trained on it
- HuggingFaceM4/idefics2-8bImage text → text 5.4K
- xtuner/llava-llama-3-8b-v1_1-transformersImage text → text 4.8K
- bczhou/tiny-llava-v1-hfImage text → text 2.6K
- xtuner/llava-llama-3-8b-v1_1-ggufImage → text 2.2K
- xtuner/llava-phi-3-mini-ggufImage → text 1.6K
- xtuner/llava-phi-3-mini-hfImage → text 556
- HuggingFaceM4/idefics2-8b-chattyImage text → text 538
- chatpig/llava-llama3Image → text 383
How to load it
from datasets import load_dataset
ds = load_dataset("Lin-Chen/ShareGPT4V")