Dataset · DeepGlint-AI
100M Chinese image-text pairs
DeepGlint-AI/DanQing100M
A massive 12 TB web corpus from 2024–2025 for pre-training multimodal models.
What's inside
A large-scale dataset containing 10 to 100 million Chinese image-text pairs collected from the web between 2024 and 2025. The total dataset size is 12 TB, intended for pre-training vision-language models.
What you can do with it
- Pre-training large multimodal models and CLIP-level encoders in Chinese
- Training zero-shot image classification and caption generation systems
- Tasks
- Zero shot image classification, Image → text
- Languages
- Chinese
- Size
- 10M–100M rows
- Downloads this month
- 1.1K
- Likes
- 52
How to load it
from datasets import load_dataset
ds = load_dataset("DeepGlint-AI/DanQing100M")