Enhanced Spatiotemporal Consistency for Image-to-LiDAR Data Pretraining
IEEE Transactions on Pattern Analysis and Machine Intelligence · 2026-03 · 48 (3) · 3819-3834
DOI: 10.1109/tpami.2025.3640589 ↗Kurz gefasst
Fachartikel in der Zeitschrift IEEE Transactions on Pattern Analysis and Machine Intelligence, 2026-03. Verfasst von Xiang Xu und 7 weiteren Autor:innen. Die Arbeit wurde bisher 2-mal zitiert (Crossref) und nennt 106 Literaturangaben. Gefördert von: National Natural Science Foundation of China, Key Research and Development Program of Jiangsu Province, Ministry of Education - Singapore.
Abstract (Originalsprache)
LiDAR representation learning has emerged as a promising approach to reducing reliance on costly and labor-intensive human annotations. While existing methods primarily focus on spatial alignment between LiDAR and camera sensors, they often overlook the temporal dynamics critical for capturing motion and scene continuity in driving scenarios. To address this limitation, we propose SuperFlow++, a novel framework that integrates spatiotemporal cues in both pretraining and downstream tasks using consecutive LiDAR-camera pairs. SuperFlow++ introduces four key components: (1) a view consistency alignment module to unify semantic information across camera views, (2) a dense-to-sparse consistency regularization mechanism to enhance feature robustness across varying point cloud densities, (3) a flow-based contrastive learning approach that models temporal relationships for improved scene understanding, and (4) a temporal voting strategy that propagates semantic information across LiDAR scans to improve prediction consistency. Extensive evaluations on 11 heterogeneous LiDAR datasets demonstrate that SuperFlow++ outperforms state-of-the-art methods across diverse tasks and driving conditions. Furthermore, by scaling both 2D and 3D backbones during pretraining, we uncover emergent properties that provide deeper insights into developing scalable 3D foundation models. With strong generalizability and computational efficiency, SuperFlow++ establishes a new benchmark for data-efficient LiDAR-based perception in autonomous driving.
Förderung
- National Natural Science Foundation of China
- Key Research and Development Program of Jiangsu Province
- Ministry of Education - Singapore
- MOE AcRF Tier 2
- NTU S-Lab and industry partne
Zitieren
APA
Xu, X., Kong, L., Shuai, H., Zhang, W., Pan, L., Chen, K., Liu, Z., & Liu, Q. (2026). Enhanced Spatiotemporal Consistency for Image-to-LiDAR Data Pretraining. IEEE Transactions on Pattern Analysis and Machine Intelligence, 48(3), 3819-3834. https://doi.org/10.1109/tpami.2025.3640589
Vancouver
Xu X, Kong L, Shuai H, Zhang W, Pan L, Chen K, et al. Enhanced Spatiotemporal Consistency for Image-to-LiDAR Data Pretraining. IEEE Transactions on Pattern Analysis and Machine Intelligence. 2026;48(3):3819-3834. doi:10.1109/tpami.2025.3640589
BibTeX
@article{Xu2026,
author = {Xu, Xiang and Kong, Lingdong and Shuai, Hui and Zhang, Wenwei and Pan, Liang and Chen, Kai and Liu, Ziwei and Liu, Qingshan},
title = {Enhanced Spatiotemporal Consistency for Image-to-LiDAR Data Pretraining},
journal = {IEEE Transactions on Pattern Analysis and Machine Intelligence},
year = {2026},
volume = {48},
number = {3},
pages = {3819--3834},
doi = {10.1109/tpami.2025.3640589}
}
EUVIMED