Add synthetic data to unit training

v1.3

Update README.md

Fix ONNX bug + fix Pillow bug

• Allow CPU execution provider for ONNX runtime
• Update debug code to use new Pillow functionality

Yolov8 model for units + more data

• Replace Yolov8 model with Yolov5 model
• Added object detection data from Jurek Elliesen to the training data

Minor bug fixes

• Remove pp-adb
• Faster screenshots

Example hotfix

Merge pull request #44 from Pbatch/all-contributors/add-ankushsethi

docs: add ankushsethi as a contributor for bug

Unit detection overhaul

• Moved to a bigger YoloV5 model
• Remapped classes ('ally_dragon' -> 'dragon', 'enemy_dragon' -> 'dragon'). This doubles the training data and halves the number of classes leading to big OD improvements.
• Designed a lightweight binary CNN to predict ally/enemy. Packaged as a small ONNX file.

Import hotfix

Hotfix for data imports

PyPI Release

Release for PyPI