Measuring the uncertainty of deep neural network results

This project was part of the Computer Vision course (IA716) for the advanced master of AI (Télécom Paris) 2020/2021.

All code, report and results are reported in the notebook and structured as follows:

• Presentation of the CamVid dataset used in the experiments
• Construction of a neural network for semantic segmentation
• Metrics for uncertainty evaluation of the predicted results
• Methods for decreasing the uncertainty of the network