CHANGELOG
CHANGELOG
---------
- (2.2.1)
- Issue#67: Fixed extraction logic for "ROI" boxes.
- (2.2.0)
- PR#60: Update image.py with channel_axis arg.
This bumps the required version of scikit-image up to 0.19.0.
- PR#56: Fix wrong valid check in TD3.
- (2.1.0)
- PR#54: add command to only extract the candidate ROIs
The package now provides the `extract_mrz_rois` script.
- (2.0.0)
- PR#50: Remove pdfminer3k, as pdfminer now supports Python 3.
The new PDFMiner does not support Python 2 any more, however. As a
consequence PassportEye now also
requires Python3.6 or above as of this release. Hence the major version
number change, otherwise
the internals are all the same.
- (1.4.1)
- Issue#34: ValueError: zero-size array to reduction operation fmin which
has no identity)
- (1.4.0)
- Issue#27: Reverted the forced use of legacy recognizer in Tesseract
- Changed error handling in mrz and evaluate_mrz scripts.
- Added --legacy parameter to mrz and evaluate_mrz scripts to explicitly
enable legacy recognizer.
- (1.3.0)
- PR#25, Issue#24: Support reading images from stream. Added dependency on
ImageIO.
- Tesseract forced to use legacy mode by default (as it seems to work
better than the LSTM one).
- Result of `read_mrz`, when not None, now also provides access to the OCR-
ed text in `mrz.aux['text']`.
- (1.1.2) Issue #9, PR#10: Pinned PyTesseract version to 0.1.7 (newer ones are
incompatible)
- (1.1.1)
- Minor packaging fixes
- PR#2 (https://github.com/konstantint/PassportEye/pull/2)
- Support for Python 3 (Issue #1)
- Issue #8 (https://github.com/konstantint/PassportEye/issues/8)