Research map / Detection and segmentation / CNN features
Scale-Equalizing Pyramid Convolution for Object Detection
Check your idea against 108,967 papers
AI-extracted summary. Written by AI from the paper's title only (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 |
|---|---|---|---|
| Replace head convolutions with PConv (ablation) | RetinaNet · COCO minival | AP: 35.7 -> 37.0 | Helped |
| Add integrated batch normalization after PConv (ablation) | RetinaNet + PConv · COCO minival | AP: 37.0 -> 37.9 | Helped |
| Add scale-equalizing deformable kernels (ablation) | RetinaNet + PConv + iBN · COCO minival | AP: 37.9 -> 39.7 | Helped |
| Use SEPC-lite rather than full SEPC (ablation) | RetinaNet · COCO minival | AP: 39.7 -> 38.8 | Hurt |
| Replace head convolutions with PConv (ablation) | FSAF · COCO minival | AP: 36.9 -> 38.7 | Helped |
| Add integrated batch normalization after PConv (ablation) | FSAF + PConv · COCO minival | AP: 38.7 -> 38.9 | Helped |
| Add scale-equalizing deformable kernels (ablation) | FSAF + PConv + iBN · COCO minival | AP: 38.9 -> 41.3 | Helped |
| Replace head convolutions with PConv (ablation) | FreeAnchor · COCO minival | AP: 38.5 -> 40.0 | Helped |
| Add integrated batch normalization after PConv (ablation) | FreeAnchor + PConv · COCO minival | AP: 40.0 -> 41.2 | Helped |
| Add scale-equalizing deformable kernels (ablation) | FreeAnchor + PConv + iBN · COCO minival | AP: 41.2 -> 42.8 | Helped |
| Replace full SEPC with SEPC-lite (ablation) | FreeAnchor · COCO minival | AP: 42.8 -> 41.7 | Hurt |
| Use shared single BN statistics across pyramid levels instead of integrated BN (ablation) | Detection head · COCO | AP | Hurt |
Closest papers by meaning
- Feature Pyramid Networks for Object Detection (CVPR 2017 · similarity 0.92)
- M2Det: A Single-Shot Object Detector Based on Multi-Level Feature Pyramid Network (AAAI 2019 · similarity 0.89)
- An Analysis of Scale Invariance in Object Detection SNIP (CVPR 2018 · similarity 0.88)
- Fine-Grained Dynamic Head for Object Detection (NeurIPS 2020 · similarity 0.87)
- FAS-Net: Construct Effective Features Adaptively for Multi-Scale Object Detection (AAAI 2020 · similarity 0.87)
- Parameter-Inverted Image Pyramid Networks (NeurIPS 2024 · similarity 0.85)
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.