delmalih/WSI-Segmentation-Patch

Whole Slide Images Segmentation using Patch UNet

WSI-Segmentation-Patch

Requirements

pip install -r requirements.txt

Step 1: Extracting patches

python extract_patches.py -i <IMAGES_FOLDER> \ # <-- required
                          -o <OUTPUT_FOLDER> \ # <-- required
                          -ps <PATCH_SIZE> \ # <-- default: 256
                          -ppi <PATCHES_PER_IMAGE> \ # <-- default: 1000
                          -p <TRAIN_PROPORTION> \ # <-- default: 0.9

Step 2: Run training

python train.py -ti <TRAIN_IMAGES_FOLDER> \ # <-- required
                -vi <VAL_IMAGES_FOLDER> \ # <-- required
                -m <MODEL_PATH> \ # <-- required
                -ps <PATCH_SIZE> \ # <-- default: 256
                -bs <BATCH_SIZE> \ # <-- default: 20
                -e <NB_EPOCHS> \ # <-- default: 10.000

Step 3: Run inference

Run on a single Whole Slide Image:

python inference.py -i <IMAGE_PATH> \ # <-- required
                    -m <MODEL_PATH> \ # <-- required
                    -o <OUTPUT_PATH> \ # <-- required
                    -ps <PATCH_SIZE> \ # <-- default: 256

Run on all WSIs:

python inference_all.py -i <IMAGES_FOLDER> \ # <-- required
                        -m <MODEL_PATH> \ # <-- required
                        -o <OUTPUT_FOLDER> \ # <-- required
                        -ps <PATCH_SIZE> \ # <-- default: 256

Link to the final presentation :

https://youtu.be/SAZHsepdNuk