README.md

metadata

title: CLIP BLIP Document Classifier
emoji: 📄
colorFrom: blue
colorTo: green
sdk: gradio
app_file: app.py
pinned: false

Zero-Shot Document Classification with CLIP and BLIP

This Space classifies document images into three classes:

• email
• resume
• scientific paper

It mirrors the approach used in your notebook:

• CLIP zero-shot classification with prompt ensembling
• BLIP classification with fallback chain: ITM -> VQA -> captioning
• side-by-side score visualization

Notes

• The first launch can be slow because model weights are downloaded.
• If BLIP cannot be loaded due to memory limits, the app still works with CLIP.
• The app tries to prepare real examples from nielsr/rvl_cdip_10_examples_per_class. If unavailable, placeholder examples are generated.

Local run

pip install -r requirements.txt
python app.py

Deploy script

python deploy_space.py --space-repo-id your_user/your_space_name

Use --private if you want a private Space.