Dataset · graph-based-captions
Graph-based captions for 10 million images
graph-based-captions/GBC10M
A large dataset of English graph annotations for CC12M images tailored for multimodal models.
What's inside
The dataset contains between 10 and 100 million image-text records created from CC12M images. Annotations are produced using the graph-based caption (GBC) paradigm in English: they connect individual image region captions into a single detailed text while preserving scene graph structure.
What you can do with it
- Training vision-language models (VLMs) for precise and detailed image description
- Training models to understand spatial relationships between objects in photos
- Tasks
- Image → text
- Languages
- English
- Size
- 10M–100M rows
- Downloads this month
- 746
- Likes
- 36
Models trained on it
- KBlueLeaf/TIPO-500M-ftText generation 18.6K
- KBlueLeaf/TIPO-200M-ft2Text generation 8.3K
- KBlueLeaf/TIPO-500MText generation 7.2K
- KBlueLeaf/TIPO-200MText generation 5K
- KBlueLeaf/TIPO-100MText generation 2K
- KBlueLeaf/TIPO-200M-ftText generation 656
- QuantFactory/TIPO-500M-GGUFText generation 634
- mradermacher/TIPO-500M-ft-i1-GGUF 583
How to load it
from datasets import load_dataset
ds = load_dataset("graph-based-captions/GBC10M")