Special Course in Deep Learning in Computer Vision

Repository for the Special Course in Deep Learning in Computer Vision @DTU

This repo will include 3 of the projects developed:

• P1_1: Image Classification: HotDog/NoHotDog
• P1_2: Object Detection: RCNN Waste Classification using the TACO dataset.
• P_2: Generative Adversarial Networks

Accessing the DTU Cluster

1. Open your terminal and write:

ssh [email protected]

2. Login with your credentials
3. Start up an interactive node using voltash

Setting up a venv

1. Create a directory in the HPC where you want your venvs to be stored mkdir venvs
2. Check the available modules to load module avail
3. Load the python module: module load python3/3.10.7
4. Create the venv: python3 -m venv NAME_VENV
5. You can now activate your environment wherever using: source NAME_VENV/bin/activate

Setting up your Jupyter notebook

Make sure you have installed it in your env. If not do

pip install notebook

1. Start it inside your interactive node

jupyter notebook --no-browser --port=40000 --ip=$HOSTNAME

2. Copy the values in your url that look like this:

n-62-20-1:40001

3. Open another terminal and write. Please note substitute USER by your username and make sure the above line is included after L8080:

ssh [email protected] -g -L8080:n-62-20-1:40001 -N

4. Enter your credentials
5. It will look as if nothing happened, but open your browser and write http://localhost:8080/tree?

Submitting a job to the HPC DTU's Cluster

Once you have this text file (the job script, let us call it submit.sh), you must submit it by typing in a terminal the command (Source: https://www.hpc.dtu.dk/?page_id=1416): bsub < submit.sh Job Statistics: bstat