Research map / Detection and segmentation / CNN features
Variational Feature Pyramid Networks
Read the paperCheck 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 |
|---|---|---|---|
| Variational pruning with correlated ARD prior (FullARD) instead of BiFPN-style fusion | FPN fusion network in Faster RCNN · COCO test-dev | AP: 0.322 vs 0.293 (BiFPN) | Helped |
| Variational pruning with independent ARD prior | FPN fusion network in Faster RCNN · COCO test-dev | AP: 0.315 vs 0.308 (PConv, best baseline) | Helped |
| Correlated ARD prior (FullARD) vs independent ARD prior (ablation) | variational FPN · COCO test-dev (Faster RCNN and Mask RCNN) | AP: +0.007 (Faster RCNN); +0.009 (Mask RCNN) | Helped |
| Probabilistic (variational ARD) pruning vs deterministic Lasso/gradient/lottery pruning to 9 connections (ablation) | FPN connections in Mask RCNN · COCO instance segmentation | AP: ARD 0.290 vs best deterministic 0.286 (Molchanov et al.) | Helped |
| Correlated ARD prior keeps full-network accuracy with reduced connections (ablation) | FPN connections in Mask RCNN · COCO instance segmentation | AP: 0.299 with 16 connections vs 0.299 with 63 (no pruning) | No clear effect |
| Random pruning of connections (to 16) instead of learned pruning (ablation) | FPN connections in Mask RCNN · COCO instance segmentation | AP: 0.222 vs 0.299 (no pruning) | Hurt |
Closest papers by meaning
- Feature Pyramid Networks for Object Detection (CVPR 2017 · similarity 0.91)
- M2Det: A Single-Shot Object Detector Based on Multi-Level Feature Pyramid Network (AAAI 2019 · similarity 0.89)
- Pyramidal Feature Shrinking for Salient Object Detection (AAAI 2021 · similarity 0.87)
- Few-Shot Object Detection via Variational Feature Aggregation (AAAI 2023 · similarity 0.86)
- AugFPN: Improving Multi-Scale Feature Learning for Object Detection (CVPR 2020 · similarity 0.86)
- You Only Look One-Level Feature (CVPR 2021 · 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.