← OCR (image → text) libraries · Reader board
docTR 1.1.0 on OCR (image → text)
190 files: text from 190, empty 0, errors 0, hangs 0, crashes 0. 190 files give text to at least one library.
🔴 Fails where another library reads text: none
🟡 Empty where another library reads text: none
⚔️ Compare: docTR vs PaddleOCR · docTR vs Tesseract
Changes between versions
Only one version tested so far. The next PyPI release is tested automatically.
Badge
For a README:
[](https://prometheus-maker.github.io/reader-board/ocr-doctr.html)Slowest files
| documents/doclaynet-patents-e5263ee691.png | 31.748 s, 1634 MB |
| documents/doclaynet-patents-f30f528620.png | 29.363 s, 1641 MB |
| documents/doclaynet-patents-8ab9f0f6d4.png | 28.11 s, 1562 MB |
| documents/doclaynet-patents-dfb912fd48.png | 27.931 s, 1607 MB |
| documents/doclaynet-patents-b9efefd640.png | 25.443 s, 1562 MB |
| documents/doclaynet-patents-7fcee0562d.png | 19.925 s, 1513 MB |
| documents/doclaynet-patents-a9a1744322.png | 17.113 s, 1562 MB |
| documents/doclaynet-patents-cf990a1c3a.png | 16.978 s, 1581 MB |
| documents/doclaynet-laws_and_regulations-31af11501c.png | 15.02 s, 1279 MB |
| documents/doclaynet-laws_and_regulations-92374cee0b.png | 13.879 s, 1402 MB |
Recipe: board/targets.py. Run: 2026-11-06T19:05:32Z, Python 3.13.18, Linux x86_64, 4 CPUs (log). Installed: Jinja2==3.1.6, MarkupSafe==3.0.4, PyYAML==6.0.3, RapidFuzz==3.14.6, anyascii==0.3.3, anyio==4.16.0, certifi==2026.10.23, click==8.5.1, filelock==4.2.9, fsspec==2026.9.0, h11==0.16.0, h5py==3.16.0, hf-xet==1.8.0, httpcore==1.0.9, httpx==0.28.1, huggingface_hub==1.33.0, idna==3.20, langdetect==1.0.9, ml_dtypes==0.6.0, mpmath==1.3.0, networkx==3.7, numpy==2.5.4, onnx==1.23.2, opencv-python==5.0.0.93, packaging==26.3, pillow==13.0.0, protobuf==7.36.2, pyclipper==1.4.0, pypdfium2==5.15.0, python-doctr==1.1.0, scipy==1.18.1, shapely==2.2.0, six==1.17.0, sympy==1.14.0, torch==2.15.0+cpu, torchvision==0.29.2+cpu, tqdm==4.70.1, typing_extensions==4.16.0, validators==0.36.0