AI Paper Atlas

Research map / Architectures and efficiency / CNN design

FaPN: Feature-Aligned Pyramid Network for Dense Image Prediction

Shihua Huang, Zhichao Lu, Ran Cheng, Cheng He

ICCV 2021 · Architectures and efficiency › CNN design · approach: Dense matching and estimation CNNs · 278 citations (OpenAlex)

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 changedWhereMeasuredEffect
add feature alignment module (FAM) to FPN (ablation)FPN (R50) · Cityscapes valmIoU: 77.4 -> 79.7Improved
add feature selection module (FSM) on top of FAM (ablation)FPN + FAM (R50) · Cityscapes valmIoU: 79.7 -> 80.0Improved
use FSM instead of SE module after FAM (ablation)FPN + FAM (R50) · Cityscapes valmIoU: 78.8 -> 80.0Improved
use FAM instead of deconvolution upsampling (ablation)FPN + FSM (R50) · Cityscapes valmIoU: 76.7 -> 80.0Improved
apply FAM before feature fusion rather than after (ablation)FaPN (R50) · Cityscapes valmIoU: 79.3 -> 80.0Improved
add an extra 3x3 conv to FPN (ablation)FPN (R50) · Cityscapes valmIoU: 77.4 -> 77.5No clear effect
increase backbone from R50 to R101FPN · Cityscapes valmIoU: 77.4 -> 78.9Improved
replace FPN with FaPNFaster R-CNN (R50) · COCO val2017AP small: 22.4 -> 24.5Improved
replace FPN with FaPNFaster R-CNN (R50) · COCO val2017AP large: 49.1 -> 49.1No clear effect
replace FPN with FaPNPointRend R50 · Cityscapes val boundary pixelsmean boundary mIoU: 55.9 -> 58.5Improved
add FAM to real-time FPN (ablation)FPN (R18) · Cityscapes valmIoU: 68.6 -> 73.8Improved
add FSM to real-time FPN + FAM (ablation)FPN + FAM (R18) · Cityscapes valmIoU: 73.8 -> 74.2Improved

Closest papers by meaning

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.

Open scime draft

Get the next scime Dispatch and a monthly field update

Free. We send a confirmation link first; every email has a one-click unsubscribe. Privacy