ternaus/onnx_model_optimizations

ONNX

1. Prepare weights: cat weights/lama-regular.onnx_split* > weights/lama-regular.onnx
2. Run inference: python src.predict.py -i images/masked_image.jpg -m images/mask.png -w weights/lama-regular.onnx

Convert to IR

mo --input_model weights/lama-regular.onnx --output_dir weights

python src/to_openvino.py