Logo image
Revolutionizing road safety: Harnessing low-power tech for accurate object detection
Thesis   Open access

Revolutionizing road safety: Harnessing low-power tech for accurate object detection

Kunwar Aniket Singh
California State University, Sacramento
Master of Science (MS), California State University, Sacramento
03/23/2026
Handle:
https://hdl.handle.net/20.500.12741/rep:13984

Abstract

Traffic safety Object detection Deep learning (Machine learning) Computer Vision
In today’s fast-moving traffic, drivers can miss or misread road signs due to glare, clutter, or split-second decisions. Many assistive systems are costly or depend on constant connectivity, making them hard to use in everyday vehicles and classrooms. This project tackles that gap with an approachable, on-device safety assistant that watches the road and highlights important signs in real time. We built a Sacramento-focused image set from iPhone 11 drive videos and photo variations, then trained using a deep learning model to recognize six common signs—left turn, U-turn, stop, pedestrian crossing, no left turn, and railway crossing. Running entirely on a low-power Raspberry Pi with a Pi Camera, the system analyzes live video, draws clear labels, and can play a brief audio cue—no internet required. The train-on-laptop, deploy-on-Pi workflow keeps costs low and setup simple. Overall, this work offers a practical, affordable path to smarter road safety: a small device, a focused dataset, and timely alerts that fit real-world driving.
pdf
SingKunwarAniket_Fall202516.91 MBDownloadView
TextProject Open Access

Metrics

1 Record Views

Details

Logo image