Reproduce “Detecting Car Speed using Object Detection and Depth Estimation: A Deep Learning Framework” on OpenResearch
Turn your favorite coding agents into research agents that review literature, develop hypotheses, and run experiments.
Universal app for Apple silicon and Intel. Requires macOS 11 or later.
About this paper
Subhasis Dasgupta, Arshi Naaz, Jayeeta Choudhury, Nancy Lahiri
A deep learning framework estimates vehicle speed from handheld video by integrating object detection, depth estimation, and regression analysis. The system achieved an R² of 0.81 and an RMSE of 2.24 using YOLOv8 and polynomial regression, enabling portable speed detection for enforcement.
Key insights
- The inclusion of depth estimation from MiDAS significantly improves speed prediction, increasing the R² value in linear models from 0.4 to approximately 0.52.
- The relationship between visual/depth features and vehicle speed is non-linear, with polynomial regression (up to third order) outperforming linear models by a notable margin.
- YOLOv8 consistently yields better performance compared to YOLOv5 and SSD for object detection within the framework, contributing to more accurate bounding boxes and feature extraction.
Summary by alphaXiv.
Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.