Camera-Based Vehicle Functions

Completed as a part of Bosch's PRIXEL program

Python PyTorch ONNX OpenCV

As part of Bosch’s PRIXEL program, our team developed a real-time system for detecting potholes and speed breakers using object-detection models.

We curated and annotated a dataset representing Indian road conditions, where speed breakers are often unmarked and can be difficult for drivers to anticipate. Using this dataset, we developed a camera-based detection pipeline designed to identify these road hazards in real time.

The project gave me practical experience in dataset preparation, annotation, model development, and applying computer vision to a real-world road-safety problem.

Detection results

A marked speed breaker detected by the model — A marked speed breaker being detected by the model.

A pothole detected by the model — Detected pothole.

A speed breaker and pothole detected together — A speed breaker and pothole being detected together.

An unmarked speed breaker detected by the model — An unmarked speed breaker being detected by the model.

Test dataset

Detection results on the first test batch — Result on the test dataset.

Detection results on the second test batch — Result on the test dataset.

Detection results on the third test batch — Result on the test dataset.