Our new X account is live! Follow @wizwand_team for updates
WorkDL logo mark

Dual-Awareness Attention for Few-Shot Object Detection

About

While recent progress has significantly boosted few-shot classification (FSC) performance, few-shot object detection (FSOD) remains challenging for modern learning systems. Existing FSOD systems follow FSC approaches, ignoring critical issues such as spatial variability and uncertain representations, and consequently result in low performance. Observing this, we propose a novel \textbf{Dual-Awareness Attention (DAnA)} mechanism that enables networks to adaptively interpret the given support images. DAnA transforms support images into \textbf{query-position-aware} (QPA) features, guiding detection networks precisely by assigning customized support information to each local region of the query. In addition, the proposed DAnA component is flexible and adaptable to multiple existing object detection frameworks. By adopting DAnA, conventional object detection networks, Faster R-CNN and RetinaNet, which are not designed explicitly for few-shot learning, reach state-of-the-art performance in FSOD tasks. In comparison with previous methods, our model significantly increases the performance by 47\% (+6.9 AP), showing remarkable ability under various evaluation settings.

Tung-I Chen, Yueh-Cheng Liu, Hung-Ting Su, Yu-Cheng Chang, Yu-Hsiang Lin, Jia-Fong Yeh, Wen-Chin Chen, Winston H. Hsu• 2021

Related benchmarks

TaskDatasetResultRank
Object DetectionMS COCO novel classes 2017 (val)
AP14.4
123
Object DetectionCOCO (novel)
AP (Novel)21.6
50
Few-shot Object DetectionMS COCO novel classes
mAP21.6
37
Object DetectionCOCO Base Categories 2017 (val)
AP32
17
Object DetectionCOCO novel categories 2014
AP21.6
15
Object DetectionCOCO Novel Categories 2014 (test)
AP11.9
15
Object DetectionCOCO Base Categories 2014 (test)
AP38.6
15
Object DetectionMS COCO (novel split)
1-Shot nAP11.9
15
Object DetectionMS-COCO 5-way (test)
SPE10
5
Showing 9 of 9 rows

Other info

Follow for update