Meta Faster R-CNN Few-shot Object Detection Based on Support Set Feature Enhancement
| 435 | 8 | 1609 |
| 下载次数 | 被引频次 | 阅读次数 |
小样本目标检测的元学习方法能快速适应少量训练样本，较好解决现有常规检测模型泛化能力不强、适应新任务速度缓慢、鲁棒性差的问题，具有较高的实际应用价值，但该方法对支持集特征利用能力不足、检测精度不高。为此，基于支持集特征增强，针对元学习SOTA算法Meta Faster R-CNN进行改进，从支持集背景抑制与目标特征增强两个角度出发，削弱与待查询目标无关的背景信息并加强支持集内部特征之间的联系，构建一种检测性能更高的小样本目标检测算法。实验结果表明：在PASCAL VOC Novel Set数据集上的元测试阶段，本文改进算法在1-shot、2-shot、3-shot、5-shot、10-shot下的平均精度均值(mAP@0.5)较原算法分别提升了0.066%、12.038%、12.289%、10.073%、9.539%;在元微调后的测试阶段，本文改进算法的mAP@0.5较原算法有所提升或基本持平；增强支持集特征能够有效提升小样本目标检测精度。
Abstract：The meta-learning method for few-shot object detection can quickly adapt to a small number of training samples, better solve the problems that the existing conventional detection model has weak generalization ability, slow adaptation to new tasks, and poor model robustness.The method has high practical application value, but it has insufficient ability to utilize the features of the support set, and the detection accuracy is not high.In this case, from the view of background attenuation of the support set and enhancement of target features and based on the support set feature enhancement, Meta Faster R-CNN,a meta-learning SOTA algorithm is improved to weaken the background information that has nothing to do with the target to be queried and to strengthen the connection between features within the support set.In this way, a few-shot object detection algorithm with a higher detection performance is constructed.The experimental results show that in the meta-testing stage on the PASCAL VOC Novel Set dataset, the mean average precision(mAP@0.5) of the improved algorithm in 1-shot, 2-shot, 3-shot, 5-shot and 10-shot is improved by 0.066%,12.038%,12.289%,10.073% and 9.539%,respectively; in the testing stage after meta-fine-tuning, the mAP@0.5 of the improved algorithm in this paper also improves or is basically the same as that of the original algorithm; the enhancement of the support set feature can effectively improve the few-shot object detection accuracy.
[1] 刘春磊，陈天恩，王聪，等.小样本目标检测研究综述[J].计算机科学与探索，2023,17(1):53-73.LIU C L,CHEN T E,WANG C,et al.Survey of few-shot object detection[J].Journal of Frontiers of Computer Science and Technology,2023,17(1):53-73.(in Chinese)
[2] 史燕燕，史殿习，乔子腾，等.小样本目标检测研究综述[J].计算机学报，2023,46(8):1753-1780.SHI Y Y,SHI D X,QIAO Z T,et al.A survey on recent advances in few-shot object detection[J].Chinese Journal of Computers,2023,46(8):1753-1780.(in Chinese)
[3] 宋云凯，吴原顼，叶蕴瑶，等.基于特征加权与融合的小样本遥感目标检测[J].软件导刊，2024,23(4):150-156.SONG Y K,WU Y X,YE Y Y,et al.Few-shot object detection on remote sensing images based on feature weighting and fusion[J].Software Guide,2024,23(4):150-156.(in Chinese)
[4] CHEN H,WANG Y L,WANG G Y,et al.LSTD:a low-shot transfer detector for object detection[J].Proceedings of the AAAI Conference on Artificial Intelligence,2018,32(1):2836-2843.
[5] WANG X,HUANG T E,DARRELL T,et al.Frustratingly simple few-shot object detection[C]//Proceedings of the 37th International Conference on Machine Learning.Vienna,Austria:PMLR,2020:9919–9928.
[6] 张振伟，郝建国，黄健，等.小样本图像目标检测研究综述[J].计算机工程与应用，2022,58(5):1-11.ZHANG Z W,HAO J G,HUANG J,et al.Review of few-shot object detection[J].Computer Engineering and Applications,2022,58(5):1-11.(in Chinese)
[7] HUANG G,LARADJI I,VAZQUEZ D,et al.A survey of self-supervised and few-shot object detection[J].IEEE Transactions on Pattern Analysis and Machine Intelligence,2023,45(4):4071-4089.
[8] ZHANG G J,LUO Z P,CUI K W,et al.Meta-DETR:image-level few-shot detection with inter-class correlation exploitation[J].IEEE Transactions on Pattern Analysis and Machine Intelligence,2023,45(11):12832-12843.
[9] HAN G X,HUANG S Y,MA J W,et al.Meta Faster R-CNN:towards accurate few-shot object detection with attentive feature alignment[J].Proceedings of the AAAI Conference on Artificial Intelligence,2022,36(1):780-789.
[10] YAN B W,LANG C B,CHENG G,et al.Understanding negative proposals in generic few-shot object detection[J].IEEE Transactions on Circuits and Systems for Video Technology,2024,34(7):5818-5829.
[11] YANG Z,WANG Y L,CHEN X Y,et al.Context-Transformer:tackling object confusion for few-shot detection[J].Proceedings of the AAAI Conference on Artificial Intelligence,2020,34(7):12653-12660.
[12] XIAO Y,LEPETIT V,MARLET R.Few-shot object detection and viewpoint estimation for objects in the wild[J].IEEE Transactions on Pattern Analysis and Machine Intelligence,2023,45(3):3090-3106.
[13] XU H H,WANG X Q,SHAO F M,et al.Few-shot object detection via sample processing[J].IEEE Access,2021,9:29207-29221.
[14] LU Y,CHEN X Y,WU Z X,et al.Decoupled metric network for single-stage few-shot object detection[J].IEEE Transactions on Cybernetics,2023,53(1):514-525.
[15] ZHAO X W,LIU X L,MA Y Q,et al.Temporal speciation network for few-shot object detection[J].IEEE Transactions on Multimedia,2023,25:8267-8278.
[16] XIA R Y,LI G Q,HUANG Z W,et al.Bi-path combination YOLO for real-time few-shot object detection[J].Pattern Recognition Letters,2023,165:91-97.
[17] CHEN T I,LIU Y C,SU H T,et al.Dual-awareness attention for few-shot object detection[J].IEEE Transactions on Multimedia,2023,25:291-301.
基本信息:
中图分类号:TP391.41;TP183
引用信息:
[1]马俊光,文峰,殷向阳.基于支持集特征增强的Meta Faster R-CNN小样本目标检测[J].沈阳理工大学学报,2025,44(02):48-54.
Citation Information:
[1]MA Junguang,WEN Feng,YIN Xiangyang.Meta Faster R-CNN Few-shot Object Detection Based on Support Set Feature Enhancement[J].沈阳理工大学学报 Journal of Shenyang Ligong University,2025,44(02):48-54.
基金信息:
国家重点研发计划“社会治理与智慧社会科技支撑”重点专项(2022YFC3302500)
2025-02-26
2025-02-26