UBA-OWDT：一种新型的开放世界目标检测网络

doi:10.3778/j.issn.1002-8331.2312-0103

摘要/Abstract

摘要： 开放世界目标检测（open world object detection，OWOD）的主要任务是检测已知类别和识别未知目标。此外，模型在下一个训练阶段中逐步学习已知新类。针对OW-DETR（open-world detection transformer）中未知类召回率偏低、密集目标与小目标漏检等问题，提出了一种UBA-OWDT（UCSO，BiStrip and AFDF of open-world detection transformer）开放世界目标检测网络。针对未知类召回率偏低的问题，对未知类评分优化（unknown class scoring optimization，UCSO），将生成的浅层类激活图与聚合类激活图融合，获取细粒度特征信息，提高未知类的目标评分，进而提升未知类的召回率；针对小目标漏检问题，将双条状注意力（spatial attention based on strip pooling and strip convolution，BiStrip）应用于输入特征图，捕获长程依赖，保留目标精确的位置信息，增强感兴趣目标的表征，以检测小目标；针对密集目标漏检问题，采用自适应特征动态融合（adaptive feature dynamic fusion，AFDF），根据目标大小和形状，获得不同的感受野，动态分配注意力权重，关注目标的重要部分，融合不同层级的特征，以检测密集目标。在OWOD数据集的实验结果表明，未知类召回率增值范围在0.7~1.5个百分点， mAP增值范围在0.6~1.2个百分点，与现有的开放世界目标检测方法相比，在召回率偏低、密集目标与小目标漏检问题上具有一定的优势。

关键词: 开放世界目标检测, 自适应特征动态融合, 未知类评分优化, 注意力机制

Abstract: The primary task of open world object detection (OWOD) is to detect known classes and identify unknown objects. In addition, the model incrementally learns novel known classes in the subsequent training phase. To address issues in OW-DETR (open-world detection transformer), such as low recall for unknown classes and missing dense and small objects, a novel open world detection network called UBA-OWDT (UCSO, BiStrip, and AFDF of open-world detection transformer) is proposed. To tackle the problem of low recall for unknown classes, an unknown class scoring optimization (UCSO) is introduced. This approach fuses generated shallow class activation maps with aggregated class activation maps to capture fine-grained feature information, thereby enhancing the object scores for unknown classes and improving their recall rate. To address the issue of missing small objects, a double strip attention mechanism known as BiStrip is applied to input feature maps. BiStrip captures long-range dependencies, preserves precise object position information, and enhances the representation of small objects for detection. To mitigate the problem of missing dense objects, an adaptive feature dynamic fusion (AFDF) approach is adopted. AFDF dynamically allocates attention weights based on object size and shape, obtaining different receptive fields, and focuses on critical parts of the object. It also fuses features from different levels to detect dense objects. Experimental results on the OWOD dataset show that the improved recall rate for unknown classes ranges from 0.7 to 1.5 percentage points, and the mean average precision (mAP) improvement falls within the range of 0.6 to 1.2?percentage points. Compared to existing methods for open world object detection, the proposed UBA-OWDT exhibits certain advantages in addressing issues related to low recall, missing dense objects, and small object detection.

Key words: open world object detection (OWOD), adaptive feature dynamic fusion (AFDF), unknown class scoring optimization (UCSO), attention mechanism

谢斌红, 唐彪, 张睿. UBA-OWDT：一种新型的开放世界目标检测网络[J]. 计算机工程与应用, 2025, 61(8): 215-225.

XIE Binhong, TANG Biao, ZHANG Rui. UBA-OWDT: Novel Network of Open World Object Detection[J]. Computer Engineering and Applications, 2025, 61(8): 215-225.

参考文献

[1] JOSEPH K J, KHAN S, KHAN F S, et al. Towards open world object detection[C]//Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition, 2021: 5830-5840.
[2] GUPTA A, NARAYAN S, JOSEPH K J, et al. OW-DETR: open?world detection transformer[C]//Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition, 2022: 9235-9244.
[3] 谢斌红, 张鹏举, 张睿. 结合Graph-FPN与稳健优化的开放世界目标检测[J]. 计算机科学与探索, 2023, 17(12): 2954-2966.
XIE B H, ZHANG P J, ZHANG R. Open world object detection combining Graph-FPN and robust optimization[J]. Journal of Frontiers of Computer Science and Technology, 2023, 17(12): 2954-2966.
[4] LIN T Y, DOLLáR P, GIRSHICK R, et al. Feature pyramid networks for object detection[C]//Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition, 2017: 2117-2125.
[5] GUO Z, LIU C, ZHANG X, et al. Beyond bounding-box: Convex-hull feature adaptation for oriented and densely packed object detection[C]//Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition, 2021: 8792-8801.
[6] 张婷, 张兴忠, 王慧民, 等. 基于图神经网络的变电站场景三维目标检测[J]. 计算机工程与应用, 2023, 59(9): 329-336.
ZHANG T, ZHANG X Z, WANG H M, et al. 3D object detection in substation scene based on graph neural network [J]. Computer Engineering and Applications, 2023, 59(9): 329-336.
[7] 黄磊, 杨媛, 杨成煜, 等. FS-YOLOv5: 轻量化红外目标检测方法[J]. 计算机工程与应用, 2023, 59(9): 215-224.
HUANG L, YANG Y, YANG C Y, et al. FS-YOLOv5: lightweight infrared rode target detection method[J]. Computer Engineering and Applications, 2023, 59(9): 215-224.
[8] REN S Q, HE K M, GIRSHICK R, et al. Faster R-CNN: towards real-time object detection with region proposal networks[J]. IEEE Transactions on Pattern Analysis & Machine Intelligence, 2017, 39(6): 1137-1149.
[9] ZHU X, SU W, LU L, et al. Deformable transformers for end-to-end object detection[C]//Proceedings of the 9th International Conference on Learning Representations, Virtual Event, 2021: 3-7.
[10] JIANG P T, ZHANG C B, HOU Q, et al. LayerCAM: exploring hierarchical class activation maps for localization[J]. IEEE Transactions on Image Processing, 2021, 30: 5875-5888.
[11] PAPANDREOU G, CHEN L C, MURPHY K P, et al. Weakly-and semi-supervised learning of a deep convolutional network for semantic image segmentation[C]//Proceedings of the IEEE International Conference on Computer Vision, 2015: 1742-1750.
[12] WOO S, PARK J, LEE J Y, et al. CBAM: convolutional block attention module[C]//Proceedings of the European Conference on Computer Vision, 2018: 3-19.
[13] HU J, SHEN L, SUN G. Squeeze-and-excitation networks[C]//Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition, 2018: 7132-7141.
[14] HOU Q, ZHANG L, CHENG M M, et al. Strip pooling: rethinking spatial pooling for scene parsing[C]//Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition, 2020: 4003-4012.
[15] ZHU X, HU H, LIN S, et al. Deformable convnets v2: more deformable, better results[C]//Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition, 2019: 9308-9316.
[16] 赵振兵, 王帆帆, 刘良帅, 等. 基于注意力特征融合YOLOv5 模型的无人机输电线路航拍图像金具检测方法[J]. 电测与仪表, 2023, 60(3): 145-152.
ZHAO Z B, WANG F F, LIU L S, et al. Hardware detection method of aerial image of UAV transmission line based on attention feature fusion YOLOv5 model[J]. Electrical Measurement and Instrumentation, 2023, 60(3): 145-152.
[17] QIN Z, ZHANG P, WU F, et al. FcaNet: frequency channel attention networks[C]//Proceedings of the IEEE/CVF International Conference on Computer Vision, 2021: 783-792.
[18] YANG Z, ZHU L, WU Y, et al. Gated channel transformation for visual recognition[C]//Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition, 2020: 11794-11803.
[19] DAI J, QI H, XIONG Y, et al. Deformable convolutional networks[C]//Proceedings of the IEEE International Conference on Computer Vision, 2017: 764-773.
[20] TORRALBA A. Contextual priming for object detection[J]. International Journal of Computer Vision, 2003, 53: 169-191.
[21] HU J, SHEN L, ALBANIE S, et al. Gather-excite: exploiting feature context in convolutional neural networks[C]//Proceedings of the 32nd International Conference on Neural Information Processing Systems, 2018: 9423-9433.
[22] KRIZHEVSKY A, SUTSKEVER I, HINTON G E. Imagenet classification with deep convolutional neural networks[J]. Communications of the ACM, 2017, 60(6): 84-90.
[23] LIN T Y, MAIRE M, BELONGIE S, et al. Microsoft COCO: common objects in context[C]//Proceedings of the 13th European Conference on Computer Vision, 2014: 740-755.
[24] EVERINGHAM M, GOOL V L, WILLIAMS C K I, et al. The pascal visual object classes (VOC) challenge[J]. International Journal of Computer Vision, 2010, 88: 303-338.
[25] HE K, ZHANG X, REN S, et al. Deep residual learning for image recognition[C]//Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition, 2016: 770-778.
[26] DENG J, DONG W, SOCHER R, et al. Imagenet: a large-scale hierarchical image database[C]//Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition, 2009: 248-255.
[27] CARON M, TOUVRON H, MISRA I, et al. Emerging properties in self-supervised vision transformers[C]//Proceedings of the IEEE/CVF International Conference on Computer Vision, 2021: 9650-9660.
[28] KINGMA D P, BA J. Adam: a method for stochastic optimization[J]. arXiv:1412.6980, 2014.
[29] SHMELKOV K, SCHMID C, ALAHARI K. Incremental learning of object detectors without catastrophic forgetting[C]//Proceedings of the IEEE International Conference on Computer Vision, 2017: 3400-3409.
[30] PENG C, ZHAO K, LOVELL B C. Faster ILOD: incremental learning for object detectors based on faster RCNN[J]. Pattern Recognition Letters, 2020, 140: 109-115.
[31] LIU S, HUANG D, WANG Y. Learning spatial fusion for single-shot object detection[J]. arXiv:1911.09516, 2019.