Research map / Detection and segmentation / CNN features
Hilbert Curve-Encoded Rotation-Equivariant Oriented Object Detector with Locality-Preserving Spatial Mapping
Qi Ming, Liuqian Wang, Juan Fang, Xudong Zhao, Yucheng Xu, Ziyi Teng, Yue Zhou, Xiaoxi Hu, Xiaohan Zhang, Yufei Guo
Read the paperCheck your idea against 137,217 papers
AI-extracted. The measured effects below were read by AI from the paper's full text (no summary yet). Numbers and claims can be wrong or missing context, so check the paper before you cite or build on it.
Measured effects reported in the paper
| What was changed | Where | Measured | Effect |
|---|---|---|---|
| add Hilbert curve traversal convolution (HCTConv) to multi-scale features (ablation) | two-stage oriented detector baseline · DOTA val | mAP (%): 67.5 -> 69.1 | Improved |
| add HPFormer cross-scale attention after HCTConv (ablation) | HERO-Det · DOTA val | mAP (%): 69.1 -> 70.2 | Improved |
| add Residual Oriented Response Network (RORN) for regression (ablation) | baseline detector · DOTA val | mAP (%): 67.5 -> 68.6 | Improved |
| add Hilbert circular shifting (HCS) for rotation-invariant classification (ablation) | RORN detector · DOTA val | mAP (%): 68.6 -> 69.2 | Improved |
| flatten features with raster scan and apply 1x9 conv instead of 3x3 conv (ablation) | baseline detector · DOTA val | mAP (%): 67.5 -> 67.2 | No clear effect |
| use Hilbert curve instead of Z-order or Peano curve for convolution scan path (ablation) | HCTConv · DOTA val | mAP (%): 68.0 (Z-order), 64.0 (Peano) -> 69.1 | Improved |
| use Peano curve scan path (ablation) | HCTConv · DOTA val | mAP (%): 67.5 -> 64.0 | Hurt |
| enlarge Hilbert conv kernel from 1x7 to 1x9 (ablation) | HCTConv · DOTA val | mAP (%): 68.8 -> 69.1 | Improved |
| replace naive ORN with RORN (dense residual, continuous angles) (ablation) | OAPH · DOTA val | mAP (%): 67.1 -> 67.5 | Improved |
| add Hilbert curve unfolding on top of RORN (ablation) | OAPH · DOTA val | mAP (%): 67.5 -> 68.6 | Improved |
| use Hilbert curve for both flattening and reconstruction instead of raster scan (ablation) | HPFormer · DOTA val | mAP (%): 67.2 -> 69.1 | Improved |
| increase HCS shifting count n from 0 to 3 (ablation) | HCS · DOTA val, bridge (BR) | AP (%): 43.1 -> 44.3 | Improved |
Earlier methods this paper compares with
Named baselines from earlier accepted papers in this paper's results tables: better than 1 of 1 on most shared results.
- FRED — FRED: Towards a Full Rotation-Equivariance in Aerial Image Object Detection (AAAI 2024) · better on 1 of 1 shared result
Matched by method name in results tables read by AI. Authors' own tables tend to favour the newer method, and a comparison can be missed when a method is renamed.
Closest papers by meaning
- FRED: Towards a Full Rotation-Equivariance in Aerial Image Object Detection (AAAI 2024 · similarity 0.86)
- ReDiffDet: Rotation-equivariant Diffusion Model for Oriented Object Detection (CVPR 2025 · similarity 0.86)
- R2Det: Exploring Relaxed Rotation Equivariance in 2D Object Detection (ICLR 2025 · similarity 0.85)
- Point2RBox: Combine Knowledge from Synthetic Visual Patterns for End-to-end Oriented Object Detection with Single Point Supervision (CVPR 2024 · similarity 0.84)
- GauCho: Gaussian Distributions with Cholesky Decomposition for Oriented Object Detection (CVPR 2025 · similarity 0.84)
- Selective Disk Bispectrum: A Complete and Rotation Invariant Image Descriptor (NeurIPS 2026 · similarity 0.83)
Writing a paper in this area? scime draft is a free LaTeX editor with AI included, real-time co-authors and a check of your references and venue rules before you submit. It also imports OpenAI Prism projects.
