potocnikvid / fri-2022-diploma

[Master's thesis] Predicting child appearance from facial images of parents using deep learning

Author: Matjaž Mav

Mentors: doc. dr. Luka Šajn and izr. prof. dr. Vitomir Štruc

Organization: University of Ljubljana, Faculty of Computer and Information Science

---

Abstract

TODO

Requirements

• Docker
• NVIDIA Container Runtime
• Conda (see ./conda_env.yml)
• NVIDIA GPU, NVIDIA drivers (see NVlabs/stylegan2-ada)
• Create copy of the .env file, name it dev.env and fill the values
• conda create -n diploma_env python=3.9.7 deepface docker-py gitpython ipykernel ipython jupyter_client jupyter_core keras keras-preprocessing matplotlib networkx notebook numpy oauthlib opencv pandas pillow pygithub py-opencv python-dotenv pytorch pytorch-lightning requests scikit-learn scipy sklearn-contrib-lightning tensorboard tensorflow torchvision torchmetrics wandb -c conda-forge
• conda create env --name diploma python=3.9.7 deepface docker-py gitpython ipykernel ipython jupyter_client jupyter_core keras keras-preprocessing matplotlib networkx notebook numpy oauthlib opencv pandas pillow pygithub py-opencv python-dotenv pytorch pytorch-lightning requests scikit-learn scipy sklearn-contrib-lightning tensorboard tensorflow torchvision torchmetrics wandb -c conda-forge
• conda create env -n diploma -c conda-forge python=3.9.7 deepface docker-py gitpython ipykernel ipython jupyter_client jupyter_core keras keras-preprocessing matplotlib networkx notebook numpy oauthlib opencv pandas pillow pygithub py-opencv python-dotenv pytorch pytorch-lightning requests scikit-learn scipy sklearn-contrib-lightning tensorboard tensorflow torchvision torchmetrics wandb

Notes

List of useful commands:

# List processes using NVIDIA GPU resources
$ sudo fuser -v /dev/nvidia*

# List NVIDIA GPU information
$ nvidia-smi

# Run python script detached
$ rm nohup.out && nohup python script.py &
$ tail -f nohup.out