Research map / Detection and segmentation / CNN features
Explicit Shape Encoding for Real-Time Instance Segmentation
Wenqiang Xu, Haiyang Wang, Fubo Qi, Cewu Lu
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 Chebyshev polynomial fitting to the IR signature (ablation) | YOLOv3 IR · Pascal VOC 2012 val | mAP^r@0.5: 48.8 -> 62.6 | Improved |
| add Chebyshev polynomial fitting to XY contour vertices (ablation) | YOLOv3 XY · Pascal VOC 2012 val | mAP^r@0.5: 46.1 -> 48.3 | Improved |
| use inner-center radius signature instead of XY vertices (both Chebyshev fitted) (ablation) | YOLOv3 Cheby · Pascal VOC 2012 val | mAP^r@0.5: 53.1 -> 62.6 | Improved |
| use Chebyshev fitting instead of Fourier series fitting (ablation) | YOLOv3 IR · Pascal VOC 2012 val | mAP^r@0.5: 37.5 -> 62.6 | Improved |
| use Chebyshev fitting instead of polynomial regression (ablation) | YOLOv3 IR · Pascal VOC 2012 val | mAP^r@0.5: 26.3 -> 62.6 | Improved |
| increase Chebyshev dimension from 20 to 40 (ablation) | YOLOv3 IR-Cheby · Pascal VOC 2012 val | mAP^r@0.5: 62.6 -> 60.7 | Hurt |
| increase Fourier dimension from 20 to 40 (ablation) | YOLOv3 IR-Fourier · Pascal VOC 2012 val | mAP^r@0.5: 37.5 -> 36.1 | Hurt |
| increase direct IR sampling from 40 to 60 points (ablation) | YOLOv3 IR · Pascal VOC 2012 val | mAP^r@0.5: 52.6 -> 51.7 | Hurt |
| pretrain on COCO | YOLOv3-Cheby (20) · Pascal VOC 2012 val | mAP^r@0.5: 62.6 -> 69.3 | Improved |
| replace YOLOv3 with RetinaNet as base detector (ablation) | ESE-Seg · Pascal VOC 2012 val | mAP^r@0.5: 62.6 -> 65.9 | Improved |
| replace YOLOv3 with YOLOv3-tiny as base detector (ablation) | ESE-Seg · Pascal VOC 2012 val | mAP^r@0.5: 62.6 -> 53.2 | Hurt |
| use explicit Chebyshev shape encoding instead of implicit embedding (YOLO base) | YOLO · Pascal SBD 2012 val | mAP^r@0.5: 34.6 -> 40.7 | Improved |
Closest papers by meaning
- SegDINO3D: 3D Instance Segmentation Empowered by Both Image-Level and Object-Level 2D Features (AAAI 2026 · similarity 0.85)
- ShapeMask: Learning to Segment Novel Objects by Refining Shape Priors (ICCV 2019 · similarity 0.85)
- OMG-Seg: Is One Model Good Enough For All Segmentation? (CVPR 2024 · similarity 0.84)
- SOLOv2: Dynamic and Fast Instance Segmentation (NeurIPS 2020 · similarity 0.84)
- FastInst: A Simple Query-Based Model for Real-Time Instance Segmentation (CVPR 2023 · similarity 0.84)
- Segment Anything, Even Occluded (CVPR 2025 · similarity 0.84)
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.
