Research map / Detection and segmentation / CNN features
Soft Anchor-Point Object Detection
Chenchen Zhu, Fangyi Chen, Zhiqiang Shen, Marios Savvides
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 soft-weighted anchor points (down-weight points near box boundaries) (ablation) | FSAF · COCO val2017 | AP: 35.9 -> 37.0 | Improved |
| add soft-weighted anchor points (ablation) | FSAF · COCO val2017 | AP75: 37.9 -> 39.5 | Improved |
| add soft-selected pyramid levels on top of soft weighting (ablation) | FSAF + SW · COCO val2017 | AP: 37.0 -> 38.0 | Improved |
| add balanced feature pyramid (BFP) to the baseline (ablation) | FSAF · COCO val2017 | AP: 35.9 -> 36.8 | Improved |
| add SW and SS training strategy on top of BFP (ablation) | FSAF + BFP · COCO val2017 | AP: 36.8 -> 38.8 | Improved |
| vary eta of generalized centerness from 1.0 to 2.0 (ablation) | SW · COCO val2017 | AP: 37.0 -> 36.6 | Hurt |
| vary top-k selected levels from 2 to 5 (ablation) | SAPD soft selection · COCO val2017 | AP: 37.9 -> 38.0 -> 37.9 -> 37.9 | No clear effect |
| train feature selection network jointly but do not use its weights (ablation) | FSAF + SW · COCO val2017 | AP: 37.0 -> 37.1 | No clear effect |
| replace test-time center-ness (cls) with training-time soft weighting (ablation) | FSAF · COCO val2017 | AP: 36.1 -> 37.0 | Improved |
| add center-ness on regression branch on top of soft weighting (ablation) | FSAF + SW · COCO val2017 | AP: 37.0 -> 36.8 | No clear effect |
| apply soft weighting to classification loss only (ablation) | FSAF · COCO val2017 | AP: 35.9 -> 33.3 | Hurt |
| apply soft weighting to localization loss only (ablation) | FSAF · COCO val2017 | AP: 35.9 -> 35.6 | Hurt |
Earlier methods this paper compares with
Named baselines from earlier accepted papers in this paper's results tables: better than 3 of 3 on most shared results.
- FreeAnchor — FreeAnchor: Learning to Match Anchors for Visual Object Detection (NeurIPS 2019) · better on 1 of 1 shared result
- FCOS — FCOS: Fully Convolutional One-Stage Object Detection (ICCV 2019) · better on 1 of 1 shared result
- Cascade R-CNN — Cascade R-CNN: Delving Into High Quality Object Detection (CVPR 2018) · 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
- Bridging the Gap Between Anchor-Based and Anchor-Free Detection via Adaptive Training Sample Selection (CVPR 2020 · similarity 0.91)
- Dynamic Anchor Learning for Arbitrary-Oriented Object Detection (AAAI 2021 · similarity 0.89)
- Dynamic Anchor Feature Selection for Single-Shot Object Detection (ICCV 2019 · similarity 0.89)
- FCOS: A Simple and Strong Anchor-free Object Detector (TPAMI 2020 · similarity 0.89)
- Feature Selective Anchor-Free Module for Single-Shot Object Detection (CVPR 2019 · similarity 0.89)
- Multiple Anchor Learning for Visual Object Detection (CVPR 2020 · similarity 0.89)
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.
