DCSF-KD: Dynamic Channel-wise Spatial Feature Knowledge Distillation for Object Detection
Proceedings of the AAAI Conference on Artificial Intelligence · 11.04.2025 · 39 (3) · 2627-2635
DOI: 10.1609/aaai.v39i3.32266 ↗Kurz gefasst
Fachartikel in der Zeitschrift Proceedings of the AAAI Conference on Artificial Intelligence, online erschienen am 11.04.2025. Verfasst von Tao Dai und 4 weiteren Autor:innen. Die Arbeit wurde bisher 14-mal zitiert (Crossref).
Abstract (Originalsprache)
Knowledge distillation (KD) has recently gained great success in the field of object detection. By transferring the knowledge of the spatial or channel domain from the teacher model to the student model, it allows for a more compact representation with minimal performance loss. Despite this progress, existing KD methods typically treat knowledge from spatial or channel domains independently, ignoring the exploitation of the mutual relationship between these domains. In this work, we first explore the connection between spatial and channel domains and find there exists a strong correlation between them, i.e. the salient channels tend to contain significant object regions in the spatial domain. Motivated by this observation, we propose DCSF-KD, a novel Dynamic Channel-wise Spatial Feature Knowledge Distillation framework for object detection by fully exploiting both spatial and channel knowledge. Specifically, we introduce channel-wise spatial feature distillation and global channel attention distillation, using information from both domains to improve the accuracy of the student network. Experiments demonstrate that our DCSF-KD outperforms existing detection methods on both homogeneous and heterogeneous teacher-student network pairs. For example, when using the MaskRCNN-Swin detector as the teacher, and based on RetinaNet and FCOS with ResNet-50 on MS COCO, our DCSF-KD can achieve 41.9% and 44.1% mAP, respectively.
Zitieren
APA
Dai, T., Lin, Y., Guo, H., Wang, J., & Zhu, Z. (2025). DCSF-KD: Dynamic Channel-wise Spatial Feature Knowledge Distillation for Object Detection. Proceedings of the AAAI Conference on Artificial Intelligence, 39(3), 2627-2635. https://doi.org/10.1609/aaai.v39i3.32266
Vancouver
Dai T, Lin Y, Guo H, Wang J, Zhu Z. DCSF-KD: Dynamic Channel-wise Spatial Feature Knowledge Distillation for Object Detection. Proceedings of the AAAI Conference on Artificial Intelligence. 2025;39(3):2627-2635. doi:10.1609/aaai.v39i3.32266
BibTeX
@article{Dai2025,
author = {Dai, Tao and Lin, Yang and Guo, Hang and Wang, Jinbao and Zhu, Zexuan},
title = {DCSF-KD: Dynamic Channel-wise Spatial Feature Knowledge Distillation for Object Detection},
journal = {Proceedings of the AAAI Conference on Artificial Intelligence},
year = {2025},
volume = {39},
number = {3},
pages = {2627--2635},
doi = {10.1609/aaai.v39i3.32266}
}
EUVIMED