Attention transfer from human to neural networks for road object detection in winter

Boisclair, J., Kelouwani, S., Ayevide, F. K., Amamou, A., Alam, M. Z. et Agbossou, K. (2022). Attention transfer from human to neural networks for road object detection in winter. IET Image Processing, 16 (13). pp. 3544-3556. ISSN 1751-9659 1751-9667 DOI 10.1049/ipr2.12562

Prévisualisation

PDF Disponible sous licence Creative Commons Attribution Non-commercial No Derivatives. Télécharger (1MB) | Prévisualisation

Résumé

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.

Type de document:	Article
Date de dépôt:	22 janv. 2026 13:30
Dernière modification:	22 janv. 2026 13:30
Version du document déposé:	Version officielle de l'éditeur
URI:	https://depot-e.uqtr.ca/id/eprint/12540

Actions (administrateurs uniquement)

Éditer la notice