Small Object Detection Based on Structure Perception and Global Context Information

doi:10.3778/j.issn.1002-8331.2302-0275

Abstract

Abstract: In the small targets detection task, it’s difficult to detect the small objects with limited pixels and unabundant characteristics. So it is easy to cause many problems, such as missed or failed detection, and low accuracy of the model. To?address?these?issues, this paper proposes a novel multi-scale structure perception and global contextual information network for small object detection. Firstly, a multi-scale structure perception module (MSSP) is proposed to capture the detailed features of small targets, it can enhance the ability of model to identify objects with different sizes. Secondly, in order to obtain more global features, a global context module (GCM) is introduced to extract global information and effectively establish the relationship between different pixels. Finally, a new IoU loss function, namely W-CIoU, is designed for the tiny objects detection, which can relieve the gradient explosion phenomenon caused by too few small target pixels during model training. Extensive experiments show that the proposed approach achieves higher accuracy than other classic lightweight methods. Compared with the baseline, the proposed model obtains over 6.4 percentage points mAP50 gain and 4.6 percentage points mAP50：95 gain on the VisDrone dataset, and also has a good performance on the TinyPerson dataset.

Key words: small object detection, attention mechanism, context information, loss function

摘要： 在小目标检测任务中，由于小目标像素值少、特征不丰富和难提取等局限性，容易导致模型漏检、误检以及精度低等问题，提出了一种基于多尺度结构感知和全局上下文信息的小目标检测算法。针对复杂场景设计了多尺度结构感知模块，可以更好地捕获小目标的细节特征，以此增强模型识别不同尺寸物体的检测能力。为了获取更多的全局特征，借助Transformer捕获长距离依赖特征的优势设计了全局上下文信息模块，有效地建立起不同区域像素点之间的联系。针对模型训练时的梯度爆炸现象，设计了一种新的带权重损失函数W-CIoU，使得训练时的收敛速度有明显改善。大量的实验结果表明，提出的方法相较于其他经典的轻量级方法取得了较好的检测效果。与基准模型相比，提出的模型在VisDrone数据集上mAP50和mAP50：95分别提高了6.4和4.6个百分点，同时在TinyPerson数据集上也有着不错的表现。

关键词: 小目标检测, 注意力机制, 上下文信息, 损失函数

LI Zhonghua, LIN Chujun, ZHU Hengliang, LIAO Shiyu, BAI Yunqi. Small Object Detection Based on Structure Perception and Global Context Information[J]. Computer Engineering and Applications, 2024, 60(9): 292-298.

李钟华, 林初俊, 朱恒亮, 廖诗宇, 白云起. 基于结构感知和全局上下文信息的小目标检测[J]. 计算机工程与应用, 2024, 60(9): 292-298.

References

[1] CAI Z, VASCONCELOS N. Cascade R-CNN: delving into high quality object detection[C]//Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition, 2018: 6154-6162.
[2] RUKHOVICH D, SOFIIUK K, GALEEV D, et al. Iterdet: iterative scheme for object detection in crowded environments[C]//Joint IAPR International Workshops on Statistical Techniques in Pattern Recognition (SPR) and Structural and Syntactic Pattern Recognition (SSPR). Cham: Springer, 2021: 344-354.
[3] GE Z, LIU S, WANG F, et al. Yolox: exceeding yolo series in 2021[J]. arXiv:2107.08430, 2021.
[4] SUN P, ZHANG R, JIANG Y, et al. Sparse R-CNN: end-to-end object detection with learnable proposals[C]//Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition, 2021: 14454-14463.
[5] DUAN K, BAI S, XIE L, et al. Centernet: keypoint triplets for object detection[C]//Proceedings of the IEEE/CVF International Conference on Computer Vision, 2019: 6569-6578.
[6] LIN T Y, DOLLAR 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.
[7] DENG C, WANG M, LIU L, et al. Extended feature pyramid network for small object detection[J]. IEEE Transactions on Multimedia, 2021, 24: 1968-1979.
[8] SUNKARA R, LUO T. No more strided convolutions or pooling: a new CNN building block for low-resolution images and small objects[J]. arXiv:2208.03641, 2022.
[9] 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.
[10] WOO S, PARK J, LEE J Y, et al. CBAM: convolutional block attention module[C]//Proceedings of the European Conference on Computer Vision (ECCV), 2018: 3-19.
[11] VASWANI A, SHAZEER N, PARMAR N, et al. Attention is all you need[C]//Advances in Neural Information Processing Systems, 2017.
[12] DOSOVITSKIY A, BEYER L, KOLESNIKOV A, et al. An image is worth 16x16 words: transformers for image recognition at scale[J]. arXiv:2010.11929, 2020.
[13] ZHENG Z, WANG P, LIU W, et al. Distance-IoU loss: faster and better learning for bounding box regression[C]//Proceedings of the AAAI Conference on Artificial Intelligence, 2020: 12993-13000.
[14] DU D, ZHU P, WEN L, et al. VisDrone-DET2019: the vision meets drone object detection in image challenge results[C]//Proceedings of the IEEE/CVF International Conference on Computer Vision Workshops, 2019.
[15] YU X, GONG Y, JIANG N, et al. Scale match for tiny person detection[C]//Proceedings of the IEEE/CVF Winter Conference on Applications of Computer Vision, 2020: 1257-1265.
[16] LAW H, DENG J. Cornernet: detecting objects as paired keypoints[C]//Proceedings of the European Conference on Computer Vision (ECCV), 2018: 734-750.
[17] TAO Y, ZONGYANG Z, JUN Z, et al. Low-altitude small-sized object detection using lightweight feature-enhanced convolutional neural network[J]. Journal of Systems Engineering and Electronics, 2021, 32(4): 841-853.
[18] ZHAO T, WEI X, YANG X. Improved YOLO v5 for railway PCCS tiny defect detection[C]//2022 14th International Conference on Advanced Computational Intelligence (ICACI), 2022: 85-90.
[19] LI C, LI L, JIANG H, et al. YOLOv6: a single-stage object detection framework for industrial applications[J]. arXiv:2209.02976, 2022.
[20] WANG C Y, BOCHKOVSKIY A, LIAO H Y M. YOLOv7: trainable bag-of-freebies sets new state-of-the-art for real-time object detectors[J]. arXiv:2207.02696, 2022.
[21] REDMON J, FARHADI A. YOLOv3: an incremental improvement[J]. arXiv:1804.02767, 2018.
[22] CHEN L C, ZHU Y, PAPANDREOU G, et al. Encoder-decoder with atrous separable convolution for semantic image segmentation[C]//Proceedings of the European Conference on Computer Vision (ECCV), 2018: 801-818.
[23] HE K, ZHANG X, REN S, et al. Spatial pyramid pooling in deep convolutional networks for visual recognition[J]. IEEE Transactions on Pattern Analysis and Machine Intelligence, 2015, 37(9): 1904-1916.
[24] LIU S, HUANG D. Receptive field block net for accurate and fast object detection[C]//Proceedings of the European Conference on Computer Vision (ECCV), 2018: 385-400.
[25] ZHANG Y F, REN W, ZHANG Z, et al. Focal and efficient IOU loss for accurate bounding box regression[J]. Neuro Computing, 2022, 506: 146-157.