TY - JOUR
T1 - Enhanced Head
T2 - Exploring Strong Detection Heads With Vision Transformer
AU - Du, Zewen
AU - Hu, Zhenjiang
AU - Zhao, Guiyu
AU - Jin, Ying
AU - Ma, Hongbin
N1 - Publisher Copyright:
© 1999-2012 IEEE.
PY - 2025
Y1 - 2025
N2 - As a crucial component of object detectors, current detection heads often lack the capability to effectively utilize contextual information, adapt to deformable objects, and align features and tasks. However, most existing methods prioritize a single capability, lacking comprehensive approaches to introduce them simultaneously. In this paper, we propose the Enhanced Head to integrate the above three capabilities into the detectors concurrently. Specifically, we propose three attention blocks with linear complexity: Global Concentrated Attention (GCA), Local Deformable Cross-Task Attention (LDCA), and Boundary-Aware Cross-Task Attention (BACA). The GCA captures long-range dependencies efficiently by employing Spatial Information Concentration (SIC). The LDCA improves feature alignment and deformation adaptability by enabling local deformable cross-task feature interactions. The BACA aligns classification features with localization results, enhancing task alignment and further improving deformation adaptability through a region-deformable interaction scheme. We implement Enhanced Head as a plug-and-play detection head and evaluate its effectiveness through extensive experiments on the MS COCO and VisDrone datasets. For instance, on the COCO detection benchmark, our Enhanced Head achieves +3.6 AP gain for FSAF, +3.3 AP for RetinaNet, and +2.9 AP for ATSS while reducing the FLOPs.
AB - As a crucial component of object detectors, current detection heads often lack the capability to effectively utilize contextual information, adapt to deformable objects, and align features and tasks. However, most existing methods prioritize a single capability, lacking comprehensive approaches to introduce them simultaneously. In this paper, we propose the Enhanced Head to integrate the above three capabilities into the detectors concurrently. Specifically, we propose three attention blocks with linear complexity: Global Concentrated Attention (GCA), Local Deformable Cross-Task Attention (LDCA), and Boundary-Aware Cross-Task Attention (BACA). The GCA captures long-range dependencies efficiently by employing Spatial Information Concentration (SIC). The LDCA improves feature alignment and deformation adaptability by enabling local deformable cross-task feature interactions. The BACA aligns classification features with localization results, enhancing task alignment and further improving deformation adaptability through a region-deformable interaction scheme. We implement Enhanced Head as a plug-and-play detection head and evaluate its effectiveness through extensive experiments on the MS COCO and VisDrone datasets. For instance, on the COCO detection benchmark, our Enhanced Head achieves +3.6 AP gain for FSAF, +3.3 AP for RetinaNet, and +2.9 AP for ATSS while reducing the FLOPs.
KW - Detection head
KW - object detection
KW - vision transformer
UR - https://www.scopus.com/pages/publications/105015079449
U2 - 10.1109/TMM.2025.3604917
DO - 10.1109/TMM.2025.3604917
M3 - Article
AN - SCOPUS:105015079449
SN - 1520-9210
VL - 27
SP - 7834
EP - 7848
JO - IEEE Transactions on Multimedia
JF - IEEE Transactions on Multimedia
ER -