Project work for the 2021 "Maschinelles Lernen für Physiker*innen" seminar at TU Dortmund by Lucas Cremer, Daniel Wall and Lukas Beiske.
Flowers-299 by Bogdan Cretu hosted on kaggle with the CC0: Public Domain License.
Please download the dataset and save it into a folder named Flowers299 in the same parent directory as the repo, if you want to run the code in this repo.
Training of the networks is mostly done by uploading notebooks to Google Colaboratory for faster computing resources.
The first three cells in the hyperparameter_tuning.ipynb notebook can only be executed in Colab and a slight change in cell 5 has to be made, if you want to run this notebook locally:
Change path = drive_path to path = local_path.