rop_classification

This repository contains a docker image for rop classification (No rop, pre-plus and plus) from vessel segmented images.

Pulling image from command line:

$ docker pull docker.pkg.github.com/qtim-lab/rop_classification/rop:v0

Using as base image in dockerfile:

FROM docker.pkg.github.com/qtim-lab/rop_classification/rop:v0

How to use the docker image:

Loading into the container will automatically set the working directory to src/ which contains all the code.

However, the dataset (segmented vessels) must be mounted while running the container.

Below is an example:

nvidia-docker run --rm -it -v {Absolute path to segmented images in local system}:/segmented -v {Path to code}:/src rop_classification:v0

Once inside the container, to load a pre-trained model and run inference, run:

cd src
python3 main.py

Note: The above script will load a pre-trained model and will allow training for 1 epoch. Further parameters can be changed by modifying main.py

Authors:

1. Ashwin Vaswani
2. Katharina Hoebel
3. Praveer Singh