Minor/Major project B.Tech
- Detection of Potholes and asphalt deformalities to alert the driver before hand
- Porject these visual alerts on the windscreen to the driver knows the exact position of the poholes
- Track the facial expressions of the driver and predict whether the driver is alert or not
We are using the following libraries for our project
- Tensor FLow API- To Train the models on the potholes dataset
- OpenCv - To capture the Video feed and do the image processing
- Dllib - Facial landmarking
- Working of the system
- Geting the Video feed from the camera or the Mobile phone.
- Use Opencv to perform the preprocessing on the input Video feed frame by frame.
- Feeding the Input to a tensorflow image classifier that predicts and gives and location of the poholes in the image.
- Plotting the Bounding box using Opencv and projecting them on the screen.
- Parallely Using another camera or phone monitoring the driver's facial expressing and alerting the driver when he/she is not paying attention to the driving and the road is busy.
- Making of the system
- Making a dataset of Pothole images and resizing them using OpenCV.
- Lbaleing these images to train our Deep learning model using tranfer learning.
- Training our CNN to recognise these potholes in the given photos.
- comparing different models to know what gives the best speed to accuracy ratio
- using this model , implementing a predictor function to identify the potholes in the give video feed
- For facial expressions repeating the above steps for detecting the distracted driver
- Making a decission script to relate the road condition and driver alertness to decide whether to alert the driver or not.