Projects developed for the course Deep Learning of the EIT Digital data science master at UPM
The first exercise aims to build a deep artificial neural network for a classification problem using the California Housing Prices dataset. The problem consists of estimating the approximate location of housing blocks.
From the second assignment, the tasks are focused on the field of Computer Vision. In the second notebook, as first approach to solve image recognition problems, we have been asked to use feed forward Neural Networks (ffNN). Despite this not being the best practice for solving an image classification problem, the work done and approach taken provided the understanding and the knowledge on how deep neural networks work. Moreover we got familiar to their architectures and parameters.
The second approach used in the third assignment was based on Convolutional Neural Networks (CNN).CNNs are one of the most used tools for images classifications nowadays. Carrying out this work, we understood the differences between ffNNs and CNNs, how CNNs work and we explored some popular architectures.
Finally, in the last assignment, the work was focused in traffic signs detection and classification, where we were provided with some pictures and we have to detect and classify the traffic signs that appears inthose pictures.
A full report on the computer vission assignments is available with each of the tasks carefully detailed.
- Student Name 1: Stefano Baggetto
- Student Name 2: Giorgio Segalla
- Student Name 3: Angel Igareta (angel@igareta.com)