davanstrien/falcon-ocr-test

Document Processing using Falcon OCR (plain mode)

This dataset contains OCR results from images in davanstrien/ufo-ColPali using Falcon OCR, a 0.3B early-fusion vision-language model.

Processing Details

• Source Dataset: davanstrien/ufo-ColPali
• Model: tiiuae/Falcon-OCR
• Task Mode: plain - Full-page text extraction
• Number of Samples: 5
• Processing Time: 2.3 min
• Processing Date: 2026-04-07 17:25 UTC

Configuration

• Image Column: image
• Dataset Split: train
• Max Output Tokens: 2,048
• Backend: Transformers

Model Information

Falcon OCR is a compact early-fusion model that processes image patches and text tokens in a shared Transformer. Key results:

• 80.3% on olmOCR benchmark
• 88.64% on OmniDocBench
• 87.1% on multi-column documents (best in class)
• 90.3% on tables (best in class)

Reproduction

uv run https://huggingface.co/datasets/uv-scripts/ocr/raw/main/falcon-ocr.py \
    davanstrien/ufo-ColPali \
    <output-dataset> \
    --task-mode plain \
    --image-column image

Generated with UV Scripts