run.bat

echo "Create test random train dataset with circles and rectangles"
build_host\train_unet_darknet2d.exe --test-dataset-create=./dataset/train/imgs/,./dataset/train/masks/,256,256,30,0,255,0,255,0,0
echo "Create test random valid dataset with circles and rectangles"
build_host\train_unet_darknet2d.exe --test-dataset-create=./dataset/valid/imgs/,./dataset/valid/masks/,256,256,30,0,255,0,255,0,0

start dataset\train\imgs
start dataset\train\masks
start dataset\valid\imgs
start dataset\valid\masks

echo "Convert exist dataset to needed size and crop"
build_host\train_unet_darknet2d.exe --convert-images="64,64,128,128,./dataset/train/imgs/,./dataset/train/converted_imgs/"
build_host\train_unet_darknet2d.exe --convert-images="64,64,128,128,./dataset/train/masks/,./dataset/train/converted_masks/"
start dataset\train\converted_imgs
start dataset\train\converted_masks

echo "Start training here we can also downscale image during the train"
build_host\train_unet_darknet2d.exe ^
 --model-darknet=./model/unet3c2cl2l8f.cfg ^
 --epochs=200 ^
 --checkpoints-output=./checkpoints_128x128_test ^
 --train-directories=./dataset/train/imgs,./dataset/train/masks ^
 --valid-directories=./dataset/valid/imgs,./dataset/valid/masks ^
 --colors-to-class-map="rectangle,0,255,0,disk,255,0,0" ^
 --selected-classes-and-thresholds="rectangle,0.3,disk,0.3" ^
 --batch-count=3 ^
 --size-downscaled=128,128 ^
 --grayscale=no