5LSM0: Neural Networks for Computer Vision

Welcome to the repository for 5LSM0: Neural Networks for Computer Vision, a course offered by the Department of Electrical Engineering at Eindhoven University of Technology. This course is hosted by the Architectures for Relaible Image Analysis Lab.

Using different models

The four different models are placed in their own folders within the "Final assignment" folder. To run a specific model, please copy its content into the model.py, predict.py, and train.py files within the "Final assignment" folder. Example: If you want to run the Residual U-Net model, copy the content of model_Residual_U_Net.py, predict_Residual_U_Net.py, and train_Residual_U_Net.py into model.py, predict.py, and train.py, respectively.

Running models

After git cloning the repository and setting up the MobaXterm environment, perform a git pull on MobaXterm to obtain the selected model. After that, go into the "Final assignment" folder and perform the following commands:

chmod +x jobscript_slurm.sh
sbatch jobscript_slurm.sh

Download the best checkpoint, place it in the "Final assignment" folder, and rename it to model.pt.

Submission

Build the docker image by using:

docker build -t nncv-submission:latest -f "Final assignment/Dockerfile" "Final assignment"

Test it locally by running (on Windows Powershell):

docker run --rm `
  -v "${PWD}\local_data:/data" `
  -v "${PWD}\local_output:/output" `
  nncv-submission:latest

Export image to .tar for submission:

docker save -o nncv_submission.tar nncv-submission:latest

After this, submit each model to both the peak performance server and the robustness server.

Usernames

The usernames used in both the peak performance and the robustness servers can be seen below:

• Danny_baseline
• Danny_Residual_U_Net
• Danny_SE_U_Net_v2
• Danny_Residual_SE_U_Net_v2

The TU/e email address can be seen below: d.h.h.huynh@student.tue.nl