AI Paper Atlas

Research map / Detection and segmentation / CNN features

Scene Parsing With Global Context Embedding

Wei-Chih Hung, Yi-Hsuan Tsai, Xiaohui Shen, Zhe Lin, Kalyan Sunkavalli, Xin Lu, Ming-Hsuan Yang

ICCV 2017 · Detection and segmentation › CNN features · approach: Local features and context in CNNs · 59 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 global context feature encoding (ablation)FCN-8s · ADE20K valMean IU: 29.39 -> 32.48Improved
add global prior encoding (ablation)FCN-8s · ADE20K valMean IU: 29.39 -> 32.92Improved
add global context feature encoding (ablation)DeepLab · ADE20K valMean IU: 33.94 -> 37.06Improved
add global prior encoding (ablation)DeepLab · ADE20K valMean IU: 33.94 -> 37.46Improved
combine feature and prior encoding with joint training (ablation)DeepLab · ADE20K valMean IU: 33.94 -> 38.37Improved
combine feature and prior instead of prior alone (ablation)DeepLab + Prior · ADE20K valPixel accuracy: 77.94 -> 77.76Hurt
add feature and prior encodingDeepLab · PASCAL Context valMean IU: 44.38 -> 46.52Improved
add feature and prior encodingDeepLab · PASCAL Context valPixel accuracy: 71.57 -> 73.80Improved
use learned global context feature instead of VGG-16 fc7 for retrieval (ablation)retrieval · ADE20K val scene retrieval (Kp=5)F2 score: 0.2524 -> 0.5601Improved
vary Ka in affinity matrix construction (ablation)global context network · ADE20K val retrievalF2 score: 0.5527 -> 0.5601No clear effect
increase Kp from 1 to 5 retrieved images (ablation)prior generation · ADE20K val retrievalF2 score: 0.4420 -> 0.5601Improved
increase Kp from 5 to 10 retrieved images (ablation)prior generation · ADE20K val retrievalF2 score: 0.5601 -> 0.5228Hurt

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