TY - GEN
T1 - GTAD
T2 - 2025 IEEE/RSJ International Conference on Intelligent Robots and Systems, IROS 2025
AU - Li, Tianhao
AU - Li, Yang
AU - Li, Mengtian
AU - Deng, Yisheng
AU - Ge, Weifeng
N1 - Publisher Copyright:
© 2025 IEEE.
PY - 2025
Y1 - 2025
N2 - Accurately perceiving dynamic environments is a fundamental task for autonomous driving and robotic systems. Existing methods inadequately utilize temporal information, relying mainly on local temporal interactions between adjacent frames and failing to leverage global sequence information effectively. To address this limitation, we investigate how to effectively aggregate global temporal features from temporal sequences, aiming to achieve occupancy representations that efficiently utilize global temporal information from historical observations. For this purpose, we propose a global temporal aggregation denoising network named GTAD, introducing a global temporal information aggregation framework as a new paradigm for holistic 3D scene understanding. Our method employs an in-model latent denoising network to aggregate local temporal features from the current moment and global temporal features from historical sequences. This approach enables the effective perception of both fine-grained temporal information from adjacent frames and global temporal patterns from historical observations. As a result, it provides a more coherent and comprehensive understanding of the environment. Extensive experiments on the nuScenes and Occ3D-nuScenes benchmark and ablation studies demonstrate the superiority of our method.
AB - Accurately perceiving dynamic environments is a fundamental task for autonomous driving and robotic systems. Existing methods inadequately utilize temporal information, relying mainly on local temporal interactions between adjacent frames and failing to leverage global sequence information effectively. To address this limitation, we investigate how to effectively aggregate global temporal features from temporal sequences, aiming to achieve occupancy representations that efficiently utilize global temporal information from historical observations. For this purpose, we propose a global temporal aggregation denoising network named GTAD, introducing a global temporal information aggregation framework as a new paradigm for holistic 3D scene understanding. Our method employs an in-model latent denoising network to aggregate local temporal features from the current moment and global temporal features from historical sequences. This approach enables the effective perception of both fine-grained temporal information from adjacent frames and global temporal patterns from historical observations. As a result, it provides a more coherent and comprehensive understanding of the environment. Extensive experiments on the nuScenes and Occ3D-nuScenes benchmark and ablation studies demonstrate the superiority of our method.
UR - https://www.scopus.com/pages/publications/105029944239
U2 - 10.1109/IROS60139.2025.11246664
DO - 10.1109/IROS60139.2025.11246664
M3 - 会议稿件
AN - SCOPUS:105029944239
T3 - IEEE International Conference on Intelligent Robots and Systems
SP - 4075
EP - 4082
BT - IROS 2025 - 2025 IEEE/RSJ International Conference on Intelligent Robots and Systems, Conference Proceedings
A2 - Laugier, Christian
A2 - Renzaglia, Alessandro
A2 - Atanasov, Nikolay
A2 - Birchfield, Stan
A2 - Cielniak, Grzegorz
A2 - De Mattos, Leonardo
A2 - Fiorini, Laura
A2 - Giguere, Philippe
A2 - Hashimoto, Kenji
A2 - Ibanez-Guzman, Javier
A2 - Kamegawa, Tetsushi
A2 - Lee, Jinoh
A2 - Loianno, Giuseppe
A2 - Luck, Kevin
A2 - Maruyama, Hisataka
A2 - Martinet, Philippe
A2 - Moradi, Hadi
A2 - Nunes, Urbano
A2 - Pettre, Julien
A2 - Pretto, Alberto
A2 - Ranzani, Tommaso
A2 - Ronnau, Arne
A2 - Rossi, Silvia
A2 - Rouse, Elliott
A2 - Ruggiero, Fabio
A2 - Simonin, Olivier
A2 - Wang, Danwei
A2 - Yang, Ming
A2 - Yoshida, Eiichi
A2 - Zhao, Huijing
PB - Institute of Electrical and Electronics Engineers Inc.
Y2 - 19 October 2025 through 25 October 2025
ER -