CS 9860 Advanced Machine Learning: Final Project / Prototype System from Nexus Labs
Western University
Overview
In my final project, I proposed making use of a person's heat signature to detect whether a room is occupied, by using cheap thermal sensors paired with a Convolutional Neural Network trained to differentiate between a person and false-positive signals found in a room, such as a hot-air vent. This project is a continuation of building a prototype energy management smart-home system from one of my previous startups, Nexus Labs. My model achieved ~99% accuracy in discriminating between true-positive and false-positive heat signatures. The motivation behind coming up with this project is that the smart-home industry has failed to produce a robust occupancy sensor that can sense stationary people in an occupied room. My project shows potential for being able to solve this problem. Watch a video of me presenting my work at the bottom of this page.
Video of My Presentation
Final Grade & Feedback
I received a grade of 100%. Feedback from my professor is copied & pasted below:
"This project proposes a very interesting problem to detect human presence from a thermal sensor. The report is very well organized. The results are very promising with proper discussion. Probably because the data collected is less varied, the deep learning model only needs to learn certain fixed patterns to distinguish. Anyhow, it is a very novel project and well established. Hope you have more further study on this."
"This project proposes a very interesting problem to detect human presence from a thermal sensor. The report is very well organized. The results are very promising with proper discussion. Probably because the data collected is less varied, the deep learning model only needs to learn certain fixed patterns to distinguish. Anyhow, it is a very novel project and well established. Hope you have more further study on this."