Dark Assistant: A Novel ML-Based Real-Time Object Recognition System for Blinds
N Harsha, Karunakara Rai B · 2023
This project focuses on developing an object identification system using Raspberry Pi, aiming to assist individuals with visual impairments in recognizing and identifying objects in their surroundings. The system leverages the capabilities of Raspberry Pi, combining it with computer vision techniques to provide real-time object recognition. The hardware setup includes a Raspberry Pi board, a camera module, and a speaker or headphones for audio feedback. The software implementation involves installing and configuring the operating system on Raspberry Pi and integrating computer vision libraries, such as OpenCV, for image processing and analysis. The object identification algorithm follows a multi-step process. Two controlled experiments are conducted where two state of the art object detection models SSD and YOLO are tested in how they perform in accuracy. Results show that the SSD model slightly outperforms YOLO in accuracy, but with the low processing power that the current generation of Raspberry Pi has to offer, none of the two performs well enough to be viable in applications where high speed is necessary. Object recognition is done by the Pre-trained model MobileNet for recognizing the object with significant accuracy. The proposed model time utilization for multiple object identification in same distance is from minimum of 16.7% to maximum of 20% in cumulative basis.