Github repository of the PhD project of Alexander Hunt at the University of Edinburgh

This repository contains the machine code for the PhD project of Alexander Hunt at the University of Edinburgh. The project is supervised by Prof. Till Bachmann and Prof. Bob Fisher.

Overview

machine-code contains the data processing pipeline, the AI interface and automatic results analysis for the PhD project. The pipeline is designed to process data imported from a microscope capturing device. The pipeline is designed to be modular and flexible, allowing for easy importing of data wether it be images or videos. The data will be converted to jpeg still images then cut into 416x416 pixel images. These images will be fed into a YOLOvX object detection model to detect present cells. The results will be analysed to determine the accuracy of the model or provide a report upon the sample. The pipeline is designed to be run on a server with a GPU to allow for fast processing of the data.

Installation

If you wish to use docker skip to the docker section.

First make sure you have the following installed:

• Python3
• pip3
• CUDA 12.0
• cuDNN 8.0.5
• OpenCV 4.5.3 (compiled with CUDA for python and installed in PATH)
• darknet compiled with CUDA, cuDNN and OpenCV and installed in PATH (https://github.com/AlexeyAB/darknet)

Then run the following commands to install the required python packages:

pip3 install -r requirements.txt

Usage

To run the program, run the following command:

python3 main.py

or on linux/mac:

./main.py

Then follow the instructions on the screen. The program will guide you through the process of importing data, processing it and analysing the results. The output path will be defined by you during the process.

Docker

requirements:

• docker
• nvidia-docker (container runtime for GPU access)

Mount the root data directory as a volume /media to the container

-v /path/to/data:/media 

Building locally

To run the program in a docker container, first make sure you have docker installed. Then run the following commands:

docker build -t machine-code .

Running the container

docker run -it --gpus all machine-code

And follow the instructions on screen.

Pulling prebuilt image

Alternatively, you can pull the prebuilt image from dockerhub:

docker pull alex1075/machine-code

Then run the container:

docker run -it --gpus all alex1075/machine-code

PS you can run the program/docker container inside a tmux session to keep it running without an active user session.