Traffic Light Detection Using Tensorflow Object Detection Framework
- 1 October 2019
- conference paper
- conference paper
- Published by Institute of Electrical and Electronics Engineers (IEEE) in 2019 IEEE 9th International Conference on System Engineering and Technology (ICSET)
Abstract
Traditional methods in machine learning for detecting traffic lights and classification are replaced by the recent enhancements of deep learning object detection methods by success of building convolutional neural networks (CNN), which is a component of deep learning. This paper presents a deep learning approach for robust detection of traffic light by comparing two object detection models and by evaluating the flexibility of the TensorFlow Object Detection Framework to solve the real-time problems. They include Single Shot Multibox Detector (SSD) MobileNet V2 and Faster-RCNN. Our experimental study shows that Faster-RCNN delivers 97.015%, which outperformed SSD by 38.806% for a model which had been trained using 441 images.Keywords
This publication has 3 references indexed in Scilit:
- Traffic light detection and recognition for autonomous vehiclesThe Journal of China Universities of Posts and Telecommunications, 2015
- Traffic light mapping and detectionPublished by Institute of Electrical and Electronics Engineers (IEEE) ,2011
- Traffic light recognition using image processing compared to learning processesPublished by Institute of Electrical and Electronics Engineers (IEEE) ,2009