Overview / Usage

This project is an internal ADAS feature with an aim to solve one of the most terrifying problems in the automotive industry, "Accidents". Nowadays, it's very common for a driver to be distracted by several small reasons, risking their lives in the long run. Autonomous Driving cars might kick in to solve such problems but let's say it takes a decade for them to function accurately. Initially, as an ADAS feature, this project aims to integrate adaptive cruise control with real-time drive'rs facial expressions.

If you're distracted while driving, feeling sleepy, too drunk to handle, caught reading a mail, wanted to grab something from the back etc, your car gives you a warning and if you don't respond, it automatically slows down or stops in the worst case. It doesn't take a lot to do production as this project mostly spices up the existing software+hardware a bit.

Methodology / Approach

The key skills are Deep Learning and Image Processing. We can place a sensor internally and do some sensor fusion to read in the facial expressions and analyze them later by means of a pre-trained deep neural net. The warnings could be pushed out in IHU's and trigger adaptive cruise control in the worst case. We just have to connect the dots and make this project a reality!

High-level programming like CUDA is appreciated to train the net using GPU'S such that it can perform fastly in real-time.

Technologies Used

Deep Learning, Python, TensorFlow, Image Processing, Sensor Fusion, CUDA