Project Description

This project uses a version of the YOLO (You Only Look Once) architecture in order to detect if a human is in front of the computer. If detected, an alarm will sound, and using TKinter, an alert pop-up mandating entry of password will be required. It has various customizable settings located in a JSON file including choice of password, choice of alarm-delay, choice of alarm, and others. You can find the Github link here.

Purpose

This project is less about creating an actual technical concept but demonstrating a basic use-case of pretrained deep-learning-based classifiers and their use in various fields. Impressively, this particular model uses an untrained version of YOLO and is extremely powerful, accurate, and fast!

Technical Content

In essence, the way YOLO is used in this project is through the inclusion of PyTorch files. We ran YOLO based on the contents of the YOLOv5 repository and got a resulting YOLOv5.pt file. We used PyTorch to analyze this file and start an instance of the image-classification system when the program opened in order to determine if a human was in front. To learn more about how YOLO works and how various deep-learning-based classifiers in the Computer Vision space function, please feel free to visit my notes in the Mathematics of Computer Vision Project.

Lessons Learned

Very many lessons were learned while making this website/blog. Here are the main ones:

I had assumed there was a lot more “knowledge” in the application of these deep-learning classifiers to typical modern-day-usage, but this knowledge is practically none.
YOLO is, by far, one of the easiest deep-learning based classifiers to use in order to detect and decipher various objects.

Author: Abhiram Kidambi
Written: 08-06-2024
Tags: PROJECT, TECH