Simplified Real-time Categorized Vehicle Counting and Speed Estimation

This paper presents a traffic monitoring system that counts, categorizes, and measures the speed of vehicles seen by a stationary camera. Designed to run on low-power hardware, reductions and optimizations on existing algorithms were implemented. The system achieved 3.7 fps using only five watts of power and a count within 10% of the true value with a detector having 95% recall and 95% precision.