Deep Learning-Based Traffic Light Classification with Model Parameter Selection

Yıldız G., Dizdaroğlu B., Yıldız D.

4th International Conference on Artificial Intelligence and Applied Mathematics in Engineering, D. Jude Hemanth,Tuncay Yigit,Utku Kose,Ugur Guvenc, Editör, Springer Nature, Zug, ss.197-217, 2023

  • Yayın Türü: Kitapta Bölüm / Araştırma Kitabı
  • Basım Tarihi: 2023
  • Yayınevi: Springer Nature
  • Basıldığı Şehir: Zug
  • Sayfa Sayıları: ss.197-217
  • Editörler: D. Jude Hemanth,Tuncay Yigit,Utku Kose,Ugur Guvenc, Editör
  • Karadeniz Teknik Üniversitesi Adresli: Evet

Özet

Considering the existence of autonomous vehicles, it is seen that many studies have been done on the traffic light classification recently. Automatic determination of traffic lights can significantly prevent traffic accidents. As the number of vehicles on the road increases daily, such a classification process becomes crucial. The classification process appears to result in higher accuracy using deep learning approaches. In this study, a deep learning-based classification process is performed for traffic lights. A convolutional neural network model with efficient parameters is proposed. Additively, hyperparameter adjustment is made. In addition to this, the effects of color spaces and input image sizes on the classification results are investigated. There are four classes of images with red, yellow, green, and off tags in the database used. When the results are examined, it is seen that the classification accuracy of over 96% is achieved.