AI Paper Atlas

Research map / Detection and segmentation / CNN features

SegNet: A Deep Convolutional Encoder-Decoder Architecture for Image Segmentation

Vijay Badrinarayanan, Alex Kendall, Roberto Cipolla

TPAMI 2017 · Detection and segmentation › CNN features · approach: Context modeling for semantic segmentation · 20,512 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
learn the upsampling instead of fixed bilinear interpolation (ablation)SegNet-Basic decoder · CamVid testmean I/U: 45.6 -> 46.3Improved
use max-pooling indices (SegNet-Basic) instead of FCN-style decoder (ablation)SegNet-Basic vs FCN-Basic · CamVid testmean I/U: 46.4 -> 46.3No clear effect
add encoder feature map addition to FCN decoder (ablation)FCN-Basic · CamVid testmean I/U: 42.7 -> 46.4Improved
add encoder feature map addition to SegNet decoder (ablation)SegNet-Basic · CamVid testmean I/U: 46.3 -> 48.0Improved
use multi-channel decoder filters instead of single-channel (ablation)SegNet-Basic · CamVid testmean I/U: 44.6 -> 46.3Improved
remove dimensionality reduction of encoder maps in FCN (ablation)FCN-Basic · CamVid testmean I/U: 46.4 -> 47.6Improved
use median frequency class balancing instead of natural frequency (ablation)SegNet-Basic · CamVid testmean I/U: 38.0 -> 46.3Improved
use median frequency class balancing instead of natural frequency (ablation)SegNet-Basic · CamVid testglobal accuracy: 84.1 -> 82.8Hurt
use pretrained VGG16 encoder and deeper SegNet instead of SegNet-BasicSegNet · CamVid testmean I/U: 46.3 -> 50.2Improved
train on larger 3.5K road scene datasetSegNet · CamVid testclass average accuracy: 65.9 -> 81.3Improved
train on larger 3.5K road scene datasetSegNet · CamVid testmean I/U: 50.2 -> 69.1Improved
train encoder-decoder layer-wise with L-BFGS instead of SGDSegNet-Basic · CamVid testclass average accuracy: 62.3 -> 62.9Improved

Later papers that compared with this method

15 later accepted papers report SegNet as a baseline in their results tables; 14 of them report a better number on most of the results they share with it.

7 more not listed.

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

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