We have seen many incidents of road accidents that have occurred due to negligence and unforeseen distractions, and it becomes particularly regretful when it is clear that applications of current technology more than able to prevent such mishaps. Hence, this project was born.
The Distracted Driver is a ML-based project that aims to record actions, and the facial micro-expressions of a vehicle driver and searches for signs of distraction.
We used jupyter notebook to implement a cascading classifier that was trained using more than 1800 samples of eyes to detect signs of drowsiness. We used Keras running on top of TensorFlow, numpy, pandas and other python modules. Most of the code was written in jupyter notebook in Windows 10.
For newbies like us applying ML on the fly in a project is difficult where each turn is equipped with errors and warnings. The cv2 module was particularly difficult to navigate due to its plethora of errors.
We are proud of learning how to correct errors as quickly as possible while learning how these faults made our programs tick.
We learnt a lot about Convolutional Neural Networks and accidentally learnt how to be more efficient in google searches. We learnt how to apply theoretical skills in practical situations.
The project is still at its infancy. It has a lot of potential to be exploited where the scope of data collection can be improved. We can collect more micro-expressions where even situations like strokes or Cardiac Arrests can be identified. With emergence of self driving cars they can be combined to potentially save thousands of lives.