readme.md

skin cancer

Install

1. Install Keras Mask-RCNN (https://drive.google.com/file/d/1Ug6697XbnRpEPjZyN5uLEEHyR27JyeVo/view?usp=sharing)
2. Create sym-link: (e.g. ln -s ~/Mask_RCNN/mrcnn mrcnn)
3. Download the pre-trained network and place it in the main directory: https://github.com/matterport/Mask_RCNN/releases/download/v2.0/mask_rcnn_coco.h5
4. Create isic2018 folder and place the following files:
   • Train data (https://isic-challenge-data.s3.amazonaws.com/2018/ISIC2018_Task1-2_Training_Input.zip)
   • Traing Ground Truth (https://isic-challenge-data.s3.amazonaws.com/2018/ISIC2018_Task1_Training_GroundTruth.zip)

Project structure

   |--- mask_rcnn
   | |--- isic2018/
   | | |--- ISIC2018_Task1-2_Training_Input
   | | |--- ISIC2018_Task1_Training_GroundTruth
   | |--- lesions.py
   | |--- mask_rcnn_coco.h5
   | |--- mrcnn/

Investigate

python lesions.py --mode investigate

Train

python lesions.py --mode train

Predict

python lesions.py --mode predict --image isic2018/ISIC2018_Task1-2_Training_Input/ISIC_0000000.jpg

OUTPUT 1:

OUTPUT 2:

OUTPUT 3: