Background
Camel-vehicle collisions are a serious road safety concern in parts of the Middle East and North Africa. Detecting camels near roads — especially at night — requires a system that can operate in real time with high accuracy.
This project built a live detection system using a custom-trained YOLOv5 model, served through a Flask web interface. It demonstrated the feasibility of using object detection for proactive road safety, and the underlying research was published as an arXiv paper.
Features
Live detection
Real-time camel detection through a browser-based interface — feed in video or camera input and get instant bounding-box predictions.
Custom YOLOv5 model
Trained on a curated dataset of camel images annotated for object detection, published on Kaggle for community use.
Web-based UI
Flask-powered interface making the detection model accessible without any local ML setup required.
Published research
The research behind this project was published on arXiv and has accumulated 12 citations, validating the approach.
Built with
Python
YOLOv5
Flask
PyTorch
OpenCV
Impact
12
Citations on arXiv
2023
Published
Open
Dataset on Kaggle