A Comparative Study between State-of-the-Art Object Detectors for Traffic Light Detection
R Gokul, A. J. Nirmal, K R Bharath, M. P. Pranesh, R. Karthika
20202020 International Conference on Emerging Trends in Information Technology and Engineering (ic-ETITE)28 citationsDOI
Abstract
In this paper, we evaluate some commonly used state-of-the-art object detectors, namely, Faster RCNN and YOLO, for traffic light detection. We choose the Bosch Small Traffic Light Dataset which is considered to be a challenging benchmark for this purpose. The model architecture, parameters are discussed in detail, and are altered to detect and classify even traffic lights that are indistinguishable to the human eye. We present the results for these optimized models along with the baseline results. Our experimental study shows that Faster RCNN model outperforms YOLO in terms of the precision obtained. However, when real-time deployment is considered, YOLO performs the best.
Topics & Concepts
Benchmark (surveying)Object detectionComputer scienceDetectorSoftware deploymentArtificial intelligenceComputer visionBaseline (sea)Object (grammar)State (computer science)Deep learningReal-time computingPattern recognition (psychology)AlgorithmTelecommunicationsCartographyGeographyOceanographyOperating systemGeologyAdvanced Neural Network ApplicationsVideo Surveillance and Tracking MethodsAutonomous Vehicle Technology and Safety