AI Paper Atlas

Research map / Detection and segmentation / CNN features

Laplacian Pyramid Reconstruction and Refinement for Semantic Segmentation

Golnaz Ghiasi, Charless C. Fowlkes

ECCV 2016 · Detection and segmentation › CNN features · approach: Context modeling for semantic segmentation · 505 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
replace upsampling of class scores with learned-basis reconstruction (10 bases)FCN-32s · PASCAL VOC val (w/o scale aug)mean IoU: 59.4% -> 64.1%Improved
compare LRR 32x reconstruction against FCN-8s with skip fusionFCN-8s · PASCAL VOC val (w/o scale aug)mean IoU: 62.7% -> 64.1%Improved
add multi-scale training augmentation (ablation)LRR 32x · PASCAL VOC valmean IoU: 64.07% -> 66.81%Improved
add dilation and erosion losses to the 32x branch (ablation)LRR-4x · PASCAL VOC val (32x branch)mean IoU: 71.2% -> 72.9%Improved
add dilation and erosion losses to the 32x branch (ablation)LRR-4x · PASCAL VOC val (multi-scale)mean IoU: 75.0% -> 76.6Improved
add multiplicative boundary masking to high-resolution branches (ablation)LRR-4x · PASCAL VOC valmean IoU: about 1%Improved
add the 4x layer without boundary masking (ablation)LRR fusion · PASCAL VOC valmean IoUNo clear effect
fuse predictions at multiple test scales (ablation)LRR-4x · PASCAL VOC valmean IoU: between 1.1% to 2.5%Improved
add fully connected CRF post-processing (ablation)LRR-4x · PASCAL VOC valmean IoU: between 0.9% to 1.4%Improved
pretrain on MS COCO in addition to VOCLRR-4x · PASCAL VOC 2011 valmean IoU: 74.6% -> 77.5%Improved
use more than 10 basis functions in reconstruction (ablation)LRR reconstruction layer · PASCAL VOC valmean IoUNo clear effect

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