Deep Learning Automobile Detection with LITE

October 1, 2024

At Hack OH/IO, my team and I developed a cutting-edge Deep Learning Model for the Honda Drive Challenge, designed to identify various types of vehicles on the road. This innovation addresses a critical challenge faced by law enforcement: the difficulty in quickly and accurately identifying vehicles, especially when it’s unclear what type of car is violating traffic laws.

To achieve this, we implemented a state-of-the-art LITE (Light Inception with Boosting Technique) deep learning model. Some of the technology that we used included Flask, Numpy, and Tkinter. Our solution was highly efficient, requiring only 9,000 training parameters while achieving 2.5x the speed of traditional, heavier models. My primary role in the project was integrating the user application interface with the deep learning model, ensuring a seamless and user-friendly experience in tandem with complex backend.

Check out the video below for a detailed walkthrough of our project! Additionally, if you would like to read the paper we designed our solution off of, it is linked below.

LITE: Light Inception with boosTing tEchniques for Time Series Classification