Real-time diagnosis of damage to car windshields


Challenge

Based on the acoustic signals recorded by several sensors integrated into the windshield of a vehicle, determine the location of potential impacts and whether they have caused damage requiring repair.

Resolution

After applying advanced techniques of feature engineering, the form of the detected acoustic signal and other characteristics are used to train a machine learning algorithm. The algorithm aims at predicting glass breaking (or not), with satisfactory precision level on the economic side.

Techno

Python/SciPy/scikit-learn/TensorFlow/Keras


This project is a real challenge calling for vast knowledge in data science as well as in material physics or signal processing.