Attention transfer from human to neural networks for road object detection in winter. Issue 13 (1st September 2022)

• Record Type: Journal Article
• Authors: Boisclair, Jonathan; Kelouwani, Sousso; Ayevide, Follivi Kloutse; Amamou, Ali; Alam, Muhammad Zeshan; Agbossou, Kodjo
• Is Part Of: IET image processing. Volume 16:Issue 13(2022)
• Page Start: 3544
• Page End: 3556
• Publication Date: 2022-09-01
• Subjects: Image processing -- Periodicals; 621.36705
• DOI: 10.1049/ipr2.12562
• Languages: English
• ISSNs: 1751-9659

Abstract

Abstract: As an essential feature of autonomous road vehicles, obstacle detection must be executed on a real‐time onboard platform with high accuracy. Cameras are still the most commonly used sensors in autonomous driving. Most detections using cameras are based on convolutional neural networks. In this regard, a recent teacher–student approach, called transfer learning, has been used to improve the neural network training process. This approach has only been used with a neural network acting as a teacher to the best of our knowledge. This paper proposes a novel way of improving training data based on attention transfer by getting the attention map from a human. The proposed method allows the dataset size reduction by 50%, which leads to up to a 60% decline in the training time. The experimental results indicate that the proposed method can enhance the F1‐score of the network by up to 10% in winter conditions.