Research map / Detection and segmentation / CNN features
Feature Pyramid Networks 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 single-scale RPN features with an FPN | RPN · COCO minival | AR^1k: 48.3 -> 56.3 | Helped |
| add FPN features to region-based Fast R-CNN | Fast R-CNN · COCO minival | AP: 31.9 -> 33.9 | Helped |
| use FPN in place of the controlled single-scale C4 detector | Faster R-CNN · COCO minival | AP: +2.3 AP | Helped |
| use FPN rather than the best listed competition single-model result | Faster R-CNN · COCO test-dev | AP: 36.2 vs 35.7; +0.5 | Helped |
| use FPN mask proposals instead of the previous best listed method, SharpMask | Segmentation proposal generation · COCO first 5k val images | AR: 39.8 -> 45.7 | Helped |
| add the top-down pathway to the bottom-up pyramid (ablation) | RPN · COCO minival | AR^1k: 49.5 -> 56.3 | Helped |
| add lateral connections to the top-down pyramid (ablation) | RPN · COCO minival | AR^1k: 46.1 -> 56.3 | Helped |
| use pyramid levels rather than only the finest P2 level (ablation) | RPN · COCO minival | AR^1k: 51.3 -> 56.3 | Helped |
| add the top-down pathway to the bottom-up pyramid (ablation) | Fast R-CNN · COCO minival | AP: 24.9 -> 33.9 | Helped |
| add lateral connections to the top-down pyramid (ablation) | Fast R-CNN · COCO minival | AP: 31.3 -> 33.9 | Helped |
| use all pyramid levels rather than only P2 (ablation) | Fast R-CNN · COCO minival | AP: 33.4 -> 33.9 | Helped |
| share RPN and Fast R-CNN features (ablation) | Faster R-CNN · COCO minival | AP: 33.9 -> 34.3 | Helped |
Closest papers by meaning
- M2Det: A Single-Shot Object Detector Based on Multi-Level Feature Pyramid Network (AAAI 2019 · similarity 0.93)
- Scale-Equalizing Pyramid Convolution for Object Detection (CVPR 2020 · similarity 0.92)
- Variational Feature Pyramid Networks (ICML 2022 · similarity 0.91)
- AugFPN: Improving Multi-Scale Feature Learning for Object Detection (CVPR 2020 · similarity 0.91)
- Fine-Grained Dynamic Head for Object Detection (NeurIPS 2020 · similarity 0.90)
- Construct Effective Geometry Aware Feature Pyramid Network for Multi-Scale Object Detection (AAAI 2022 · similarity 0.90)
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.