Attention Model for Extracting Saliency Map in Driving Videos

Aksoy E., YAZICI A.

28th Signal Processing and Communications Applications Conference (SIU), ELECTR NETWORK, 5 - 07 October 2020 identifier identifier

  • Publication Type: Conference Paper / Full Text
  • Doi Number: 10.1109/siu49456.2020.9302267
  • Country: ELECTR NETWORK
  • Keywords: attention mechanism, driving, computer vision, saliency, VISUAL-ATTENTION
  • Eskisehir Osmangazi University Affiliated: Yes

Abstract

Visual perception is the most important input for driving decisions. In this study a novel attention model presented for attending salient features in visual scene, that can be used for autonomous driving or advanced driver assistance systems. Proposed attention model constructs a holistic approach for driving task. The output of the model can be used as input to driving decision systems. Proposed model is a deep neural network model that inputs extracted features to an RNN model with attention mechanism. Proposed model trained and evaluated on state of the art driving attention datasets BDD-A, DR(eye)VE and general saliency dataset CAT2000.