README.md

Signet Ring Cell Detection Challenge task 1

Installation

The code to process WSI jpeg images is located at challenge folder. Thus, The procedure to analise WSI jpeg images is the following:

1. Follow the installation instructions from README.md and README_SIGNET_RING_DETECTION.md.

2. (Optional step if not using the docker container) Make sure to also install the requirements of this project (we do recommend using a virtual environment)

   cd .. pip install -r requirements.txt

3. Go to challenge folder, make a copy of the settings file and rename it as settings.py

4. Open the settings.py file and modify it as necessary.

5. Optionally, open the config/yolov3_eval_digestpath.cfg file and modify as necessary the confidence threshold (CONFTHRE), non maximum suppression threshold (NMSTHRE) and image size (IMGSIZE) from TEST section.

6. Create an folder called input at project's root and place there the WSI jpeg images to be analysed.

7. Run the Segmentation.py file to get the predictions at the output folder in the project's directory.

   python Segmentation.py

Quick test

Using provided checkpoint and sample images

1. Rename input_sample folder.

   cp -r input_sample input

2. Get sample checkpoint (if not already done). Download it at the same folder were input_sample is located.

   git clone [email protected]:giussepi/PyTorch_YOLOv3_sample_checkpoint.git

3. Rename PyTorch_YOLOv3_sample_checkpoint folder (if not already done).

   mv PyTorch_YOLOv3_sample_checkpoint checkpoints

4. Run the Segmentation.py.

   python Segmentation.py

5. Open and review the output folder