DisCo DETR: Distance-aware Multi-view Contrastive Learning for DETR Pre-training
Proceedings of the AAAI Conference on Artificial Intelligence · 14.03.2026 · 40 (29) · 24656-24664
DOI: 10.1609/aaai.v40i29.39650 ↗Kurz gefasst
Fachartikel in der Zeitschrift Proceedings of the AAAI Conference on Artificial Intelligence, online erschienen am 14.03.2026. Verfasst von Chao Ouyang und 6 weiteren Autor:innen.
Abstract (Originalsprache)
Recent self-supervised pre-training methods for object detection often rely on generic object proposals for localization and semantic feature learning for classification, but they yield limited improvements when applied to Detection Transformers (DETR) due to a lack of architectural alignment. Hence, we propose an elegant and versatile self-supervised framework tailored for DETR-like models called Distance-aware Multi-view Contrastive Learning (DisCo DETR). DisCo DETR enhances localization and semantic features through two core components. (i) Distance-aware Multi-view Object Query Fusion explicitly guides object queries to focus on spatially close objects across views, stabilizing training and improving localization accuracy. (ii) Contrastive Learning for DETR uses native bipartite matching to identify positive output pairs across views and pull them closer, enhancing semantic features discrimination with no extra matching. DisCo DETR can be seamlessly integrated into DETR-like models and achieves SOTA transfer performance on PASCAL VOC and COCO benchmarks across multiple variants.
Zitieren
APA
Ouyang, C., Bai, Y., Zhang, J., Gao, T., Hao, J., Kong, L., & Gao, D. W. (2026). DisCo DETR: Distance-aware Multi-view Contrastive Learning for DETR Pre-training. Proceedings of the AAAI Conference on Artificial Intelligence, 40(29), 24656-24664. https://doi.org/10.1609/aaai.v40i29.39650
Vancouver
Ouyang C, Bai Y, Zhang J, Gao T, Hao J, Kong L, et al. DisCo DETR: Distance-aware Multi-view Contrastive Learning for DETR Pre-training. Proceedings of the AAAI Conference on Artificial Intelligence. 2026;40(29):24656-24664. doi:10.1609/aaai.v40i29.39650
BibTeX
@article{Ouyang2026,
author = {Ouyang, Chao and Bai, Yuyang and Zhang, Jun and Gao, Tianlu and Hao, Jun and Kong, Lijun and Gao, David Wenzhong},
title = {DisCo DETR: Distance-aware Multi-view Contrastive Learning for DETR Pre-training},
journal = {Proceedings of the AAAI Conference on Artificial Intelligence},
year = {2026},
volume = {40},
number = {29},
pages = {24656--24664},
doi = {10.1609/aaai.v40i29.39650}
}
EUVIMED