Beyond Bounding-Box: Convex-Hull Feature Adaptation for Oriented and Densely Packed Object Detection
Zonghao Guo, Chang Liu, Xiaosong Zhang, Jianbin Jiao, Xiangyang Ji, Qixiang Ye
Abstract
Detecting oriented and densely packed objects remains challenging for spatial feature aliasing caused by the intersection of reception fields between objects. In this paper, we propose a convex-hull feature adaptation (CFA) approach for configuring convolutional features in accordance with oriented and densely packed object layouts. CFA is rooted in convex-hull feature representation, which defines a set of dynamically predicted feature points guided by the convex intersection over union (CIoU) to bound the extent of objects. CFA pursues optimal feature assignment by constructing convex-hull sets and dynamically splitting positive or negative convex-hulls. By simultaneously considering overlapping convex-hulls and objects and penalizing convex-hulls shared by multiple objects, CFA alleviates spatial feature aliasing towards optimal feature adaptation. Experiments on DOTA and SKU110K-R datasets show that CFA significantly outperforms the baseline approach, achieving new state-of-the-art detection performance. Code is available at github.com/SDL-GuoZonghao/BeyondBoundingBox.
Ask Lune
Know this paper in minutes, not an afternoon.
Lune can read every page of it and the top-tier papers that cite it, and gives you its claims, its evidence and its limits, each one cited to the paper it comes from. No skimming, no AI slop.
18 top-tier papers have cited it since. Know what they found before your reviewers do.
Lune can draw on
the full paper · the top-tier papers that cite it · related work
Free to start. No credit card required.
Lineage
- H2RBox-v2: Incorporating Symmetry for Boosting Horizontal Box Supervised Oriented Object Detection cites SCRDet: Towards More Robust Detection for Small, Cluttered and Rotated Objects
- H2RBox-v2: Incorporating Symmetry for Boosting Horizontal Box Supervised Oriented Object Detection cites The KFIoU Loss for Rotated Object Detection
- H2RBox-v2: Incorporating Symmetry for Boosting Horizontal Box Supervised Oriented Object Detection cites RepPoints: Point Set Representation for Object Detection
- H2RBox-v2: Incorporating Symmetry for Boosting Horizontal Box Supervised Oriented Object Detection cites Oriented RepPoints for Aerial Object Detection
- The KFIoU Loss for Rotated Object Detection cites SCRDet: Towards More Robust Detection for Small, Cluttered and Rotated Objects
- The KFIoU Loss for Rotated Object Detection cites R3Det: Refined Single-Stage Detector with Feature Refinement for Rotating Object
- Large Selective Kernel Network for Remote Sensing Object Detection cites SCRDet: Towards More Robust Detection for Small, Cluttered and Rotated Objects
- Large Selective Kernel Network for Remote Sensing Object Detection cites The KFIoU Loss for Rotated Object Detection
- Large Selective Kernel Network for Remote Sensing Object Detection cites R3Det: Refined Single-Stage Detector with Feature Refinement for Rotating Object
- Adaptive Rotated Convolution for Rotated Object Detection cites SCRDet: Towards More Robust Detection for Small, Cluttered and Rotated Objects
- Adaptive Rotated Convolution for Rotated Object Detection cites The KFIoU Loss for Rotated Object Detection
- Adaptive Rotated Convolution for Rotated Object Detection cites Oriented RepPoints for Aerial Object Detection
- R3Det: Refined Single-Stage Detector with Feature Refinement for Rotating Object cites SCRDet: Towards More Robust Detection for Small, Cluttered and Rotated Objects
- Oriented RepPoints for Aerial Object Detection cites SCRDet: Towards More Robust Detection for Small, Cluttered and Rotated Objects
Cited by top-tier papers18
- Large Selective Kernel Network for Remote Sensing Object DetectionYuxuan Li, Qibin Hou, Zhaohui Zheng, Ming-Ming Cheng et al.ICCV 2023 · 535 citations
- Oriented RepPoints for Aerial Object DetectionWentong Li, Yijie Chen, Kaixuan Hu, Jianke ZhuCVPR 2022 · 487 citations
- Adaptive Rotated Convolution for Rotated Object DetectionYifan Pu, Yiru Wang, Zhuofan Xia, Yizeng Han et al.ICCV 2023 · 154 citations
- H2RBox-v2: Incorporating Symmetry for Boosting Horizontal Box Supervised Oriented Object DetectionYi Yu, Xue Yang, Qingyun Li, Yue Zhou et al.NeurIPS 2023 · 89 citations
- The KFIoU Loss for Rotated Object DetectionXue Yang, Yue Zhou, Gefan Zhang, Jirui Yang et al.ICLR 2023 · 89 citations
Builds on8
- R3Det: Refined Single-Stage Detector with Feature Refinement for Rotating ObjectXue Yang, Junchi Yan, Ziming Feng, Tao HeAAAI 2021 · 1,109 citations
- RepPoints: Point Set Representation for Object DetectionZe Yang, Shaohui Liu, Han Hu, Liwei Wang et al.ICCV 2019 · 1,056 citations
- Objects365: A Large-Scale, High-Quality Dataset for Object DetectionShuai Shao, Zeming Li, Tianyuan Zhang, Chao Peng et al.ICCV 2019 · 1,018 citations
- SCRDet: Towards More Robust Detection for Small, Cluttered and Rotated ObjectsXue Yang, Jirui Yang, Junchi Yan, Yue Zhang et al.ICCV 2019 · 865 citations
- CARAFE: Content-Aware ReAssembly of FEaturesJiaqi Wang, Kai Chen, Rui Xu, Ziwei Liu et al.ICCV 2019 · 842 citations
Related papers
- Rethinking Boundary Discontinuity Problem for Oriented Object DetectionHang Xu, Xinyuan Liu, Haonan Xu, Yike Ma et al.CVPR 2024 · 30 citations
- CIA-SSD: Confident IoU-Aware Single-Stage Object Detector From Point CloudWu Zheng, Weiliang Tang, Sijin Chen, Li Jiang et al.AAAI 2021 · 335 citations
- Deep Dive into Gradients: Better Optimization for 3D Object Detection with Gradient-Corrected IoU SupervisionQi Ming, Lingjuan Miao, Zhe Ma, Lin Zhao et al.CVPR 2023
- FRED: Towards a Full Rotation-Equivariance in Aerial Image Object DetectionChanho Lee, Jinsu Son, Hyounguk Shon, Yunho Jeon et al.AAAI 2024 · 30 citations
- Distance-IoU Loss: Faster and Better Learning for Bounding Box RegressionZhaohui Zheng, Ping Wang, Wei Liu, Jinze Li et al.AAAI 2020 · 4,823 citations
