Note: docTR 0.12.0 requires python >= 3.10 Note: docTR 0.12.0 requires either TensorFlow >= 2.15.0 or PyTorch >= 2.0.0

[!WARNING] TensorFlow Backend Deprecation Notice

Using docTR with TensorFlow as a backend is deprecated and will be removed in the next major release (v1.0.0). We recommend switching to the PyTorch backend, which is more actively maintained and supports the latest features and models. Alternatively, you can use OnnxTR, which does not require TensorFlow or PyTorch.

This decision was made based on several considerations:

• Allows better focus on improving the core library
• Frees up resources to develop new features faster
• Enables more targeted optimizations with PyTorch

[!WARNING] This release is the last minor release supporting TensorFlow as backend

What's changed

New features

• A new lightweight recognition model viptr_tiny was added
• New built-in dataset added - COCO-Text V2

• A new custom model loading interface

  :::python

  NEW

  model = vitstr_small(pretrained=False, pretrained_backbone=False) model.from_pretrained("<PATH_TO>") # local path or url to .pt or .h5

  Instead of depending on the backend

  reco_params = torch.load('<path_to_pt>', map_location="cpu") reco_model.load_state_dict(reco_params)

  Or with TensorFlow

  reco_model.load_weights(..)

What's Changed

Breaking Changes 🛠

• [Feat] Simplify and unify model loading - from_pretrained by @felixdittrich92 in https://github.com/mindee/doctr/pull/1915
• [Build] Add TensorFlow deprecation warnings by @felixdittrich92 in https://github.com/mindee/doctr/pull/1948

New Features

• [datasets] COCO-Text V2 integration by @sarjil77 in https://github.com/mindee/doctr/pull/1888
• [references] Recognition - Allow built-in datasets usage by @sarjil77 in https://github.com/mindee/doctr/pull/1904
• [Feat] PyTorch - VIP backbone and VIPTR recognition module by @lkosh in https://github.com/mindee/doctr/pull/1912

Bug Fixes

• [Fix] Duplicated forward call in rec train scripts by @felixT2K in https://github.com/mindee/doctr/pull/1862
• [Fix] Fix invalid hOCR format & PDF/A compatiblity with the kie preditor by @felixdittrich92 in https://github.com/mindee/doctr/pull/1870
• [CI/CD] Fix conda dependency issue by @felixT2K in https://github.com/mindee/doctr/pull/1937
• [Fix] Fix merge of short strings by @Razlaw in https://github.com/mindee/doctr/pull/1947
• [bug] fix missing import by @felixT2K in https://github.com/mindee/doctr/pull/1958
• Fix: Update min_loss only on improved validation loss to ensure best … by @sneakybatman in https://github.com/mindee/doctr/pull/1961

Improvements

• [Docs] Fix typos by @jk4e in https://github.com/mindee/doctr/pull/1885
• added russian vocab by @Madhavi258 in https://github.com/mindee/doctr/pull/1902
• [FIXED] Add proper Hebrew Unicode chars vocab by @johnlockejrr in https://github.com/mindee/doctr/pull/1909
• [classification] Add PyTorch pretrained VIP checkpoints by @felixdittrich92 in https://github.com/mindee/doctr/pull/1920
• Use OpenCV instead of Shapely to compute area & length by @lachesis in https://github.com/mindee/doctr/pull/1922
• [datasets] feat: add croatian vocabulary by @cyanic-selkie in https://github.com/mindee/doctr/pull/1923
• [docs] Fix VIPTR docstring by @felixT2K in https://github.com/mindee/doctr/pull/1924
• Migrating training scripts to torchrun by @lkosh in https://github.com/mindee/doctr/pull/1933
• [datasets] Massively extend the pre-defined vocabs by @felixdittrich92 in https://github.com/mindee/doctr/pull/1928
• [datasets] Update indic & arabic & other vocabs by @felixdittrich92 in https://github.com/mindee/doctr/pull/1941
• [Fix] Improve merge for recognition of long words (#1936) by @Razlaw in https://github.com/mindee/doctr/pull/1939
• [docs] Add tools section by @felixdittrich92 in https://github.com/mindee/doctr/pull/1950
• Give instructions on how to enable GPU acceleration on apple silicon by @ulfaslak in https://github.com/mindee/doctr/pull/1955
• [models] Add viptr tiny checkpoint by @felixdittrich92 in https://github.com/mindee/doctr/pull/1963

Miscellaneous

• [misc] post release v0.11.0 & public docker cron job adjustment by @felixdittrich92 in https://github.com/mindee/doctr/pull/1860
• [misc] Avoid tf onnx test to fail the CI & mypy fixes by @felixdittrich92 in https://github.com/mindee/doctr/pull/1903
• [Docs] Fix typo in docstring by @jk4e in https://github.com/mindee/doctr/pull/1927
• [misc] Fix mypy & merge test by @felixT2K in https://github.com/mindee/doctr/pull/1945
• [docs] Add new community model (arabic) to docs by @felixT2K in https://github.com/mindee/doctr/pull/1956
• [models] Drop viptr base config by @felixT2K in https://github.com/mindee/doctr/pull/1962
• [misc] Increase minor version v0.12 by @felixT2K in https://github.com/mindee/doctr/pull/1964

New Contributors

• @jk4e made their first contribution in https://github.com/mindee/doctr/pull/1885
• @Madhavi258 made their first contribution in https://github.com/mindee/doctr/pull/1902
• @johnlockejrr made their first contribution in https://github.com/mindee/doctr/pull/1909
• @sebastianMindee made their first contribution in https://github.com/mindee/doctr/pull/1917
• @lkosh made their first contribution in https://github.com/mindee/doctr/pull/1912
• @lachesis made their first contribution in https://github.com/mindee/doctr/pull/1922
• @cyanic-selkie made their first contribution in https://github.com/mindee/doctr/pull/1923
• @ulfaslak made their first contribution in https://github.com/mindee/doctr/pull/1955
• @sneakybatman made their first contribution in https://github.com/mindee/doctr/pull/1961

Full Changelog: https://github.com/mindee/doctr/compare/v0.11.0...v0.12.0