Traffic Analysis and AI Powered Fuel Stations

The Requirement:

Fuel Stations want to keep track of the customers that come to their fuel station through their CCTV cameras,

• Count the number of vehicles of each category
• Record the license plate numbers of every vehicle and write to database
• Track the amount of time a vehicle spends at the fuel station
• Track if the Vechile is being fueled

How did we solve it? (contibution in paranthesis)

• (50%) Passed the CCTV footage to a Computer Vision pipeline that detects vehicles, the location of their license plates, and the license plates using You Only Look Once (YOLO) based object detectors
• (50%) Tracked vechiles using deep learning based tracker to keep track of vehicles
• (80%) Tracked persons in the fuel station uniform and also the fuel nozzle to identify the time spent to fuel the vehicle
• (0%) Wrote the records into the database
• (40%) Provided different versions of the codes for servers, NVIDIA Jetson boards

Technologies:

• Programming Languages: Python
• Database: MongoDB
• ML Models: PyTorch (train), TensorRT (deployment), OpenCV

Team:

• Aditya Lohia, Shivam Swarnkar, Abhimanyu Bellam, Vineet Suryan

A demo can be found here.