SCRDet: Towards More Robust Detection for Small, Cluttered and Rotated Objects

By Xue Yang, Jirui Yang, Junchi Yan, Yue Zhang, Tengfei Zhang, Zhi Guo, Xian Sun, Kun Fu. IEEE International Conference on Computer Vision, 2018

887 citations recorded by the source corpus.

Object detection has been a building block in computer vision. Though considerable progress has been made, there still exist challenges for objects with small size, arbitrary direction, and dense distribution. Apart from natural images, such issues are especially pronounced for aerial images of great importance. This paper presents a novel multi-category rotation detector for small, cluttered and rotated objects, namely SCRDet. Specifically, a sampling fusion network is devised which fuses multi-layer feature with effective anchor sampling, to improve the sensitivity to small objects. Meanwhile, the supervised pixel attention network and the channel attention network are jointly explored for small and cluttered object detection by suppressing the noise and highlighting the objects feature. For more accurate rotation estimation, the IoU constant factor is added to the smooth L1 loss to address the boundary problem for the rotating bounding box. Extensive experiments on two remote sensing public datasets DOTA, NWPU VHR-10 as well as natural image datasets COCO, VOC2007 and scene text data ICDAR2015 show the state-of-the-art performance of our detector. The code and models will be available at https://github.com/DetectionTeamUCAS.

Subfields: Computer Vision and Pattern Recognition

View Paper: https://www.semanticscholar.org/paper/456b8ae5dd6f8316c1fda46c5a8b4204c10ae320

How researchers see this paper

The researchers whose trusted networks rank it highest, strongest advocate first. Every researcher gets a different order.

In Computer Vision and Pattern Recognition

1. Luc Van Gool
2. Junwei Han
3. Dacheng Tao
4. Qi Tian
5. Ming-Hsuan Yang
6. Junchi Yan
7. Yi Yang
8. Xiaogang Wang

Citations in the corpus

• Dynamic Refinement Network for Oriented and Densely Packed Object Detection (2020, background): “SCRDet [42] added an IOU constant factor to the L1 loss term to address the boundary issue for oriented bounding boxes.”
• SCRDet++: Detecting Small, Cluttered and Rotated Objects via Instance-Level Feature Denoising and Rotation Loss Smoothing (2020, methodology): “We show that existing aerial image rotation detectors, including FADet [27], SCRDet [30] and CAD-Det [25], often use the simple attention mechanism to re-weight the output, which can be reduced into the following general form:”
• Localization Distillation for Object Detection (2022, methodology): “To address the boundary discontinuity and square-like problems, SCRDet [74] and RSDet [50] propose IoU-smooth L1 loss and modulated loss respectively for attaining smoother boundary loss, and CSL [72] proposes to use angle classification instead of angle regression.”
• Anchor-Free Oriented Proposal Generator for Object Detection (2021, methodology): “As an extended branch of object detection, oriented detection has made great progress [1]–[3], [21]–[28] in recent years.”
• Rethinking Rotated Object Detection with Gaussian Wasserstein Distance Loss (2021, methodology): “For aerial images, ICN [2], ROI-Transformer [8], SCRDet [69] and Gliding Vertex [62] are two-stage representative methods whose pipeline comprises of object localization and classification, while DRN [41], R(3)Det [67] and RSDet [42] are singlestage methods.”
• MMRotate: A Rotated Object Detection Benchmark using PyTorch (2022, methodology): “In light of this, deep learning has also been applied to areas such as faces detection [29], text detection [13, 14, 18, 22, 47] and aerial images detection [34, 35, 38, 40].”
• Center-based 3D Object Detection and Tracking (2020, background): “One solution might be to classify a different template (anchor) for each object orientation [56, 57], but this unnecessarily increases the computational burden and may introduce a large number of potential false-positive detections.”
• PIoU Loss: Towards Accurate Oriented Object Detection in Complex Environments (2020, methodology): “owded objects can be better detected and classied. Existing OBB-based approaches are mostly built on anchor-based frameworks by introducing an additional angle dimension optimized by a distance loss [24,18,19,6,41,43] on the parameter tuple (c x;c y;w;h;). While OBB has been primarily used for simpl”
• Learning Modulated Loss for Rotated Object Detection (2019, methodology): “…i) traditional deep learning meth-ods, such as Faster RCNN [29] and RetinaNet [35]; ii) scene text detection methods, like R 2 CNN [20] and RRPN [21]; iii) recently published methods for multi-category rotation detectors, includes ICN [43], RoI Transformer [44], SCRDet [45] and R 3 Det [10].”
• R3Det: Refined Single-Stage Detector with Feature Refinement for Rotating Object (2019, background): “ICN [2], ROI-Transformer [8] and SCRDet [36], achieves the most advanced performance available today.”
• Tiny Object Detection in Aerial Images (2021, background): “Unlike object detectors in natural scenes, most of the studies use anchor-based detectors to detect objects in aerial images, such as R-P-Faster R-CNN [35], YOLT [36], RoI Transformer [37], SCRDet [38],R2-CNN [4], and Mask OBB [2].”
• Learning High-Precision Bounding Box for Rotated Object Detection via Kullback-Leibler Divergence (2021, methodology): “Table 6 compares the six peer techniques, including IoU-Smooth L1 Loss [3], Modulated loss [44], RIL [33], CSL [4, 46], DCL [45], and GWD [5] on DOTA-v1.”
• Object Detection in Aerial Images: A Large-Scale Benchmark and Challenges (2021, methodology): “Therefore, these methods [18], [24], [25] use rotated NMS (R-NMS), which requires precise detections to address this problem.”
• Arbitrary-Oriented Object Detection with Circular Smooth Label (2020, methodology): “tion-over-Union (IoU) drop, resulting in inaccurate object detection, especially in case of large aspect ratios. There have been some works addressing the boundary problem. For example, IoU-smooth L1 [44] loss introduces the IoU factor, and modular rotation loss [30] increases the boundary constrain”
• Dense Label Encoding for Boundary Discontinuity Free Rotation Detection (2020, methodology): “Angle prediction based on regression often introduces boundary discontinuity [51, 30, 48, 49], mainly including periodicity of angle (PoA) and exchangeability of edges (EoE).”
• Object Detection in Optical Remote Sensing Images: A Survey and A New Benchmark (2019, methodology): “This dataset has been widely used in the earth observation community (Cheng et al., 2014; Cheng et al., 2018b; Farooq et al., 2017; Guo et al., 2018; Han et al., 2017a; Li et al., 2018; Yang et al., 2018b; Yang et al., 2017; Zhong et al., 2018).”