A Comparative Attention Framework for Better Few-Shot Object Detection on Aerial Images
Not on the map. Why?
Abstract (arXiv)
Few-Shot Object Detection (FSOD) methods are mainly designed and evaluated on natural image datasets such as Pascal VOC and MS COCO. However, it is not clear whether the best methods for natural images are also the best for aerial images. Furthermore, direct comparison of performance between FSOD methods is difficult due to the wide variety of detection frameworks and training strategies. Therefore, we propose a benchmarking framework that provides a flexible environment to implement and compare attention-based FSOD methods. The proposed framework focuses on attention mechanisms and is divided into three modules: spatial alignment, global attention, and fusion layer. To remain competitive with existing methods, which often leverage complex training, we propose new augmentation techniques designed for object detection. Using this framework, several FSOD methods are reimplemented and compared. This comparison highlights two distinct performance regimes on aerial and natural images: FSOD performs worse on aerial images. Our experiments suggest that small objects, which are harder to detect in the few-shot setting, account for the poor performance. Finally, we develop a novel multiscale alignment method, Cross-Scales Query-Support Alignment (XQSA) for FSOD, to improve the detection of small objects. XQSA outperforms the state-of-the-art significantly on DOTA and DIOR.
Key facts. Cited by 1 paper in the map (data through 14 Oct 2026). Builds on 48 of its 69 references that are in the map. First posted on arXiv 25 Oct 2022.
Builds on
48 of 69 references.
References · 48
- 2017Attention Is All You Needcited by 60,361
- 2020An Image is Worth 16x16 Words: Transformers for Image Recognition at Scalecited by 30,186
- 2014Microsoft COCO: Common Objects in Contextcited by 17,608
- 2015Faster R-CNN: Towards Real-Time Object Detection with Region Proposal Networkscited by 9,639
- 2020End-to-End Object Detection with Transformerscited by 7,414
- 2016Layer Normalizationcited by 5,669
- 2016Feature Pyramid Networks for Object Detectioncited by 5,036
- 2017Model-Agnostic Meta-Learning for Fast Adaptation of Deep Networkscited by 4,897
- 2015You Only Look Once: Unified, Real-Time Object Detectioncited by 4,852
- 2017Squeeze-and-Excitation Networkscited by 4,789
- 2016Overcoming catastrophic forgetting in neural networkscited by 4,787
- 2017Prototypical Networks for Few-shot Learningcited by 3,015
36 more
- 2017Non-local Neural Networkscited by 2,293
- 2019FCOS: Fully Convolutional One-Stage Object Detectioncited by 1,577
- 2017DOTA: A Large-scale Dataset for Object Detection in Aerial Imagescited by 623
- 2019Generalizing from a Few Examples: A Survey on Few-Shot Learningcited by 463
- 2019Object Detection in Optical Remote Sensing Images: A Survey and A New Benchmarkcited by 340
- 2020Frustratingly Simple Few-Shot Object Detectioncited by 282
- 2018Few-shot Object Detection via Feature Reweightingcited by 265
- 2019Few-Shot Object Detection with Attention-RPN and Multi-Relation Detectorcited by 206
- 2021Dynamic Head: Unifying Object Detection Heads with Attentionscited by 187
- 2019Meta R-CNN : Towards General Solver for Instance-level Few-shot Learningcited by 182
- 2020CrossTransformers: spatially-aware few-shot transfercited by 174
- 2021FSCE: Few-Shot Object Detection via Contrastive Proposal Encodingcited by 168
- 2020Few-Shot Object Detection and Viewpoint Estimation for Objects in the Wildcited by 142
- 2020Multi-Scale Positive Sample Refinement for Few-Shot Object Detectioncited by 116
- 2016Learning feed-forward one-shot learnerscited by 93
- 2021Generalized Few-Shot Object Detection without Forgettingcited by 88
- 2018RepMet: Representative-based metric learning for classification and one-shot object detectioncited by 85
- 2018LSTD: A Low-Shot Transfer Detector for Object Detectioncited by 77
- 2021DETReg: Unsupervised Pretraining with Region Priors for Object Detectioncited by 76
- 2019One-Shot Object Detection with Co-Attention and Co-Excitationcited by 67
- 2022Few-Shot Object Detection with Fully Cross-Transformercited by 64
- 2021Meta Faster R-CNN: Towards Accurate Few-Shot Object Detection with Attentive Feature Alignmentcited by 57
- 2021Hallucination Improves Few-Shot Object Detectioncited by 50
- 2021Incremental Few-Shot Instance Segmentationcited by 49
- 2021Meta-DETR: Image-Level Few-Shot Object Detection with Inter-Class Correlation Exploitationcited by 44
- 2021Universal-Prototype Enhancing for Few-Shot Object Detectioncited by 43
- 2021Few-Shot Object Detection via Association and DIscriminationcited by 39
- 2021A Survey of Self-Supervised and Few-Shot Object Detectioncited by 35
- 2020Restoring Negative Information in Few-Shot Object Detectioncited by 28
- 2020Few-shot Object Detection on Remote Sensing Imagescited by 23
- 2020Few-shot Object Detection with Self-adaptive Attention Network for Remote Sensing Imagescited by 14
- 2020Few-Shot Object Detection via Knowledge Transfercited by 13
- 2020One-Shot Object Detection without Fine-Tuningcited by 11
- 2021Dynamic Relevance Learning for Few-Shot Object Detectioncited by 5
- 2021SIMPL: Generating Synthetic Overhead Imagery to Address Zero-shot and Few-Shot Detection Problemscited by 4
- 2021Experience feedback using Representation Learning for Few-Shot Object Detection on Aerial Imagescited by 3
Built on by
Citing papers per year: 2023: 1.
1 paper.