This project identifies the number and type of vehicles from a real-time video feed and then estimates the vehicular pollution levels using the data collected. I have just posted the initial work here, the project is currently in development.
You can find the implementation in the jupyter notebook here: https://github.com/vkasojhaa/Pollution-Prediction-Tensorflow/blob/master/object_detection/pollution_prediction.ipynb
The project is based on Tensorflow Object Detection API: https://github.com/tensorflow/models/tree/master/research/object_detection
For installation refer to: https://github.com/tensorflow/models/blob/master/research/object_detection/g3doc/installation.md
If you have any queries or suggestions you can mail me: vkasojhaa@gmail.com