Overview / Usage

I am a final year undergraduate student at BITS Pilani, Pilani campus. As a part of my college thesis project, I aim to implement a model that efficiently predicts the vehicle density maps and is also able to count the number of vehicles. this would be useful for traffic estimation at different times of day at different locations. Moreover, I aim to make the model invariant to camera motion ie, my model would still be able to predict the density maps even when the camera with which the video or photo is taken, is moving.

Methodology / Approach

I would primarily be using libraries such as keras, openCV, numpy and various other libraries. I would be experimenting with building different forms of CNN models. There are currently existing models which solve my problem such as countingCNN and hydraCNN. but my aim is to use these model architectures as baseline and improve upon them. I also plan to use the Intel technologies such as OpenVINO. Getting access to the Intel AI Dev Cloud would be a boon for me as I am limited to using a small training data owing to the lack of memory on my system and low computation power.