The Three Caballeros - Deep Learning project

Project for Deep Learning course by University of Helsinki, spring 2020.

Goal of the project

The goal of the project was to train a neural network to do multi-label image classification. The group was given a training dataset of 20 000 images of resolution 128x128. The images were originally from a photo-sharing site and released under Creative Commons-licenses allowing sharing. The results were evaluated on a separate test set withheld from the group.

Repository content

The most important parts of the repository are:

• Final report of the project
• The final Jupyter notebook containing the python code for the whole pipeline of our model. The code can also be found as plain python code in the submission folder.

Other parts of the repo:

• The submission folder contains our project submission as a zip file and the content as separate files
• The documentation folder contains the final report, a todo-list and some other files related to the final report and evaluation of our models
• The notebooks folder contains different versions of the Jupyter Notebooks (Python 3) that we used during the project
• The results folder contains classification reports, confusion matrices and learning curves for different models we trained and evaluated. These models are described in the final report.
• The scripts folder contains the script used by the course staff to evaluate the submissions

Contributors

• Mikko Kotola
• Eeva-Maria Laiho
• Aki Rehn