fast-plate-ocr 1.0.0

What's Changed

Added

• New XS and S CCT model variants, trained with x3 more data (220k global plates).
• Complete rework of docs, documenting much more everything.
• Inference now works smoothly with different onnxruntime variants like onnxruntime-gpu, onnxruntime-openvino, etc.
• Support for building and customizing CCT (Compact Convolutional Transformer) models from YAML configs.
• New model building logic, allows users to build custom-based architectures while validating it with Pydantic.
• New metric val_plate_len_acc.
• Added support for categorical focal loss.
• Added more test coverage (configs, train scripts, etc.).
• New validate_dataset.py script to help check datasets before training.
• Export script now officially supports more formats like TFLite and CoreML.
• New plate config support: keep_aspect_ratio, interpolation, image_color_mode and padding_color.
• New default augmentation for RGB image mode.
• New default models, trained with much more data.
• Added examples and more docs.

Changed

• Visualize augmentation script now respects config-based preprocessing.
• Improved plate config validation.
• ONNXPlateRecognizer is now called LicensePlateRecognizer.

Full Changelog: v0.3.0...v1.0.0

fast-plate-ocr 0.3.0

What's Changed

• Add more CI/CD workflows by @ankandrew in #25
• Add more CLI train options by @ankandrew in #27
• New Global model using MobileViTV2 trained with data from +65 countries 🚀 by @ankandrew in #28

Full Changelog: v0.2.0...v0.3.0

fast-plate-ocr 0.2.0

What's Changed

• New European model using MobileViTV2 - trained on +40 countries 🚀 .
• New models and Improved logging by @ankandrew in #19 ⭐.

New Contributors

• @ankandrew made their first contribution in #19

Full Changelog: v0.1.6...v0.2.0

fast-plate-ocr 0.1.6

Full Changelog: v0.1.5...v0.1.6

Added

• Add new Argentinian model trained with more (synthetic) data.
• Add option to visualize only predictions which have low char prob.
• Add onnxsim for simplifying ONNX model when exporting.

fast-plate-ocr 0.1.5

Full Changelog: v0.1.3...v0.1.5

fast-plate-ocr 0.1.3

Full Changelog: v0.1.2...v0.1.3

fast-plate-ocr 0.1.2

Full Changelog: arg-plates...v0.1.2

HUB

Trained models HUB and related assets (.keras and .onnx).

Additional datasets

@biodatasciencearg for contributing arg_plate_dataset_biodatasciencearg.zip dataset.