The original paper is in English. Non-English content has been machine-translated and may contain typographical errors or mistranslations. ex. Some numerals are expressed as "XNUMX".
Copyrights notice
The original paper is in English. Non-English content has been machine-translated and may contain typographical errors or mistranslations. Copyrights notice
Le suivi du regard consiste à estimer où un observateur regarde à l'intérieur d'une scène. Les informations sur l'observateur et sur la scène doivent être apprises pour déterminer les directions et les points de regard. Récemment, de nombreux travaux existants se sont concentrés uniquement sur des scènes ou des observateurs. En revanche, les cadres révélés pour le suivi du regard sont limités. Dans cet article, une méthode de suivi du regard utilisant un transformateur hybride est proposée. Sur la base de la méthode conventionnelle (GazeFollow), nous menons trois développements. Tout d’abord, un transformateur hybride est appliqué pour apprendre les images de la tête et les positions du regard. Deuxièmement, la fonction de perte du flipper est utilisée pour contrôler l’erreur du point de regard. Enfin, une nouvelle couche ReLU avec le mécanisme de renaissance (reborn ReLU) est réalisée pour remplacer les couches ReLU traditionnelles à différentes étapes du réseau. Pour tester les performances de nos développements, nous entraînons notre framework développé avec l'ensemble de données DL Gaze et évaluons le modèle sur notre ensemble collecté. Grâce à nos résultats expérimentaux, il peut être prouvé que notre cadre peut obtenir de meilleures performances que nos méthodes référencées.
Jingzhao DAI
Nanjing University
Ming LI
Nanjing University
Xuejiao HU
Nanjing University
Yang LI
Nanjing University
Sidan DU
Nanjing University
The copyright of the original papers published on this site belongs to IEICE. Unauthorized use of the original or translated papers is prohibited. See IEICE Provisions on Copyright for details.
Copier
Jingzhao DAI, Ming LI, Xuejiao HU, Yang LI, Sidan DU, "GazeFollowTR: A Method of Gaze Following with Reborn Mechanism" in IEICE TRANSACTIONS on Fundamentals,
vol. E106-A, no. 6, pp. 938-946, June 2023, doi: 10.1587/transfun.2022EAP1068.
Abstract: Gaze following is the task of estimating where an observer is looking inside a scene. Both the observer and scene information must be learned to determine the gaze directions and gaze points. Recently, many existing works have only focused on scenes or observers. In contrast, revealed frameworks for gaze following are limited. In this paper, a gaze following method using a hybrid transformer is proposed. Based on the conventional method (GazeFollow), we conduct three developments. First, a hybrid transformer is applied for learning head images and gaze positions. Second, the pinball loss function is utilized to control the gaze point error. Finally, a novel ReLU layer with the reborn mechanism (reborn ReLU) is conducted to replace traditional ReLU layers in different network stages. To test the performance of our developments, we train our developed framework with the DL Gaze dataset and evaluate the model on our collected set. Through our experimental results, it can be proven that our framework can achieve outperformance over our referred methods.
URL: https://global.ieice.org/en_transactions/fundamentals/10.1587/transfun.2022EAP1068/_p
Copier
@ARTICLE{e106-a_6_938,
author={Jingzhao DAI, Ming LI, Xuejiao HU, Yang LI, Sidan DU, },
journal={IEICE TRANSACTIONS on Fundamentals},
title={GazeFollowTR: A Method of Gaze Following with Reborn Mechanism},
year={2023},
volume={E106-A},
number={6},
pages={938-946},
abstract={Gaze following is the task of estimating where an observer is looking inside a scene. Both the observer and scene information must be learned to determine the gaze directions and gaze points. Recently, many existing works have only focused on scenes or observers. In contrast, revealed frameworks for gaze following are limited. In this paper, a gaze following method using a hybrid transformer is proposed. Based on the conventional method (GazeFollow), we conduct three developments. First, a hybrid transformer is applied for learning head images and gaze positions. Second, the pinball loss function is utilized to control the gaze point error. Finally, a novel ReLU layer with the reborn mechanism (reborn ReLU) is conducted to replace traditional ReLU layers in different network stages. To test the performance of our developments, we train our developed framework with the DL Gaze dataset and evaluate the model on our collected set. Through our experimental results, it can be proven that our framework can achieve outperformance over our referred methods.},
keywords={},
doi={10.1587/transfun.2022EAP1068},
ISSN={1745-1337},
month={June},}
Copier
TY - JOUR
TI - GazeFollowTR: A Method of Gaze Following with Reborn Mechanism
T2 - IEICE TRANSACTIONS on Fundamentals
SP - 938
EP - 946
AU - Jingzhao DAI
AU - Ming LI
AU - Xuejiao HU
AU - Yang LI
AU - Sidan DU
PY - 2023
DO - 10.1587/transfun.2022EAP1068
JO - IEICE TRANSACTIONS on Fundamentals
SN - 1745-1337
VL - E106-A
IS - 6
JA - IEICE TRANSACTIONS on Fundamentals
Y1 - June 2023
AB - Gaze following is the task of estimating where an observer is looking inside a scene. Both the observer and scene information must be learned to determine the gaze directions and gaze points. Recently, many existing works have only focused on scenes or observers. In contrast, revealed frameworks for gaze following are limited. In this paper, a gaze following method using a hybrid transformer is proposed. Based on the conventional method (GazeFollow), we conduct three developments. First, a hybrid transformer is applied for learning head images and gaze positions. Second, the pinball loss function is utilized to control the gaze point error. Finally, a novel ReLU layer with the reborn mechanism (reborn ReLU) is conducted to replace traditional ReLU layers in different network stages. To test the performance of our developments, we train our developed framework with the DL Gaze dataset and evaluate the model on our collected set. Through our experimental results, it can be proven that our framework can achieve outperformance over our referred methods.
ER -