This repository presents the source code for the construction and execution of a trained agent that aims to detect images produced by a failed RGB camera.

The repository includes the source code to:

1. create a dataset of images using the Carla simulator
2. create images as if they are produced by a failed camera (for this purpose, the following tool is used, that is described in the Thesis from Francesco Secci)
3. train an agent to distinguish between "failed" and "normal" images
4. test the agent on the different images set.

All details are reported in the Thesis of Pietro Bernabei.

Link to the github:

https://github.com/BernabeiPietro/Carla_Detector_Cam_Malfunction