02456_project - Structure Tensor Hyper-resolution

This GitHub repo contains all code used for our project in the DTU course 02456 - Deep Learning.

The repository structure is as follows:

exploration_ipynbs contains jupyter notebooks that were used to develop code related to data processing and model configuration/training, which could then be used to construct .py and .sh scripts to use on DTU's HPC.

runners contains the .sh scripts we ran on HPC. The scripts differ only slightly in terms of what trainer script they use.

trainers contains the .py scripts that contain our whole pipeline from data processing/transformation to model creation, training, validation, testing, and saving of the model.

• We will in the future revisit this script to divide and refactorise into separate scripts in order to allow for easier reproducibility and maintainability, etc.

utilities contains minor utility scripts and a requirements.txt file that contains the required packages to run a trainer script.

wandb contains folders and files only relevant for wandb, and these are therefore sort of a blackbox.

Repo tree:

├── exploration_ipynbs
│   ├── DLProject.ipynb
│   ├── LoadModel.ipynb
├── runners
│   ├── submit_train_test.sh
│   ├── submit_train_test_run.sh
├── trainers
│   ├── train_script.py
│   ├── train_script_run.py
├── utilities
│   ├── image_cutter.py
│   ├── model2.py
│   ├── rename-files.py
│   ├── requirements.txt
├── wandb
│   ├── ...
├── .gitignore
└── README.md

Data is accesible on Google drive: https://drive.google.com/drive/folders/1Nl0ThLbXHwq59G8rAe5zgxeDpUX8T3nu?usp=share_link