Deep Learning for Computer Vision

Repo structure

This repo takes advantage of two frameworks: (1) Hydra for configs management and (2) Pytorch Lightning for improving our lives when colaborating and running experiments on different hardware.

The particular approach of this repo is heavily inspired by [https://youtu.be/w10WrRA-6uI].

Getting started

On HPC:

1. Create a virtual environment the way you are used to (conda, venv, pyenv, whatever). The intended approach is to use venv. If conda or other virtual environment frameworks are used, the scripts in /bash/ folder cannot be used but are easily modified. (1 line modification).

It's very important that you load a new-ish python version before running the getting started script. Do this by:

module load python3/3.10
module load cuda/11.7

2. Set environment variable:

export PATH_TO_VENV=#path to your venv

3. Run the bash script from the root folder:

./bash/setup.sh

This will:

1. Load the HPC modules mentioned above
2. Activate the virtual environment
3. Install the package in editable mode with all requirements

You should further install the requirements for developing (writing code) of the package:

pip install -r requirements-dev.txt

Contribution guide

This repo has protection on the main branch. Therefore any contribution has to go through a Pull Request.

Training

The package logs relevant metrics and stats to wandb. For this to run, one needs to login with a token:

wandb login your_token

After which we can tain on GPU:

python3 ./scripts/train.py trainer.accelerator=gpu

Or cpu:

python3 ./scripts/train.py

Training HotDog NotHotDog:

python3 scripts/train.py -cn config_hotdog_training trainer.accelerator=gpu model=resnet18_transfer model.params.num_classes=1 metric.metric.params.task=binary general.run_name=xxxxxxx

Training Segmentation Models:

python3 scripts/train.py -cn config_ph2_training trainer.accelerator=gpu general.run_name=xxxxxxx  general.project_name=project2 model=unet