To provide a system that will actually help in the day to day activities of the bravehearts that is the Police Officers out there.
What it does
It uses different technologies like Python, OpenCV, a ML Classifier and a Tiny YoloV3 model along with a beautiful GUI created in PyQt5 to provide a system that helps officers detect traffic on the roads, detect the number plates using images and live object detection and also tag and track objects/vehicles using a tracker built in Python.
How we built it
Started off creating a GUI and then working on the the other parts of the project. Collected the dataset and annotated it using a tool called LabelImg to create .txt files as the labels. Trained the model using a batch size of 1 for about 150,000 iterations for the entirety of the Hack. Used the Computer Vision Library to create a connect between the models and the code to be executed!
Challenges we ran into
How do you make sure all the processes simultaneously? How do you make an effective tracker? Which tracker do we use? The fast one or the accurate one?
Accomplishments that we're proud of
A good database is being created which will help us create a predictive model for future use as to which roads are the most crowded at different times to toggle those particular signal. A very good Neural Network model which gives out good tradeoff between accuracy and precision on both training and testing data sets!
What we learned
Basics of ML, DL. The vast area of Computer Vision. A sneak peek into what lies in the future for such technologies.
What's next for Police.AI- A smart system for the bravehearts out there!
A good predictive models for all the traffic signals to toggle traffic signals and maintain a smooth flow of traffic, provide more effective tracking mechanisms and perform live number plate detection.(Preferably on a small device like a Raspberry Pi)