AI Paper Atlas

Research map / Detection and segmentation / CNN features

Dynamic Context Correspondence Network for Semantic Alignment

Shuaiyi Huang, Qiuyue Wang, Songyang Zhang, Shipeng Yan, Xuming He

ICCV 2019 · Detection and segmentation › CNN features · approach: Local features and context in CNNs · 88 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 spatial context encoder (kernel 25) (ablation)NC-Net · PF-PASCALPCK alpha=0.1 (%): 78.9 -> 80.9Improved
add dynamic fusion with correlation embedding (no aux loss) on top of context encoder (ablation)Baseline+S · PF-PASCALPCK alpha=0.1 (%): 80.9 -> 81.0No clear effect
add multi-auxiliary task loss (ablation)Baseline+SD · PF-PASCALPCK alpha=0.1 (%): 81.0 -> 82.3Improved
use dynamic attention fusion instead of average fusion (ablation)Baseline+SAA · PF-PASCALPCK alpha=0.1 (%): 80.2 -> 82.3Improved
add correlation map embedding to dynamic fusion (ablation)Baseline+SCA · PF-PASCALPCK alpha=0.1 (%): 79.9 -> 82.3Improved
average fusion of correlation maps instead of using context feature alone (ablation)Baseline+S · PF-PASCALPCK alpha=0.1 (%): 80.9 -> 80.2Hurt
increase context kernel size from 11 to 25 (ablation)Baseline+S · PF-PASCALPCK alpha=0.1 (%): 78.9 -> 80.9Improved
increase context kernel size from 25 to 31 (ablation)Baseline+S · PF-PASCALPCK alpha=0.1 (%): 80.9 -> 77.1Hurt
add dynamic fusion and aux loss at kernel size 31 (ablation)Baseline+S · PF-PASCALPCK alpha=0.1 (%): 77.1 -> 80.7Improved
overall DCCNet vs NC-NetNC-Net · PF-PASCALPCK alpha=0.1 (%): 78.9 -> 82.3Improved
DCCNet vs NC-Net at stricter thresholdNC-Net · PF-WILLOWPCK alpha=0.05 (%): 44.0 -> 43.6Hurt
DCCNet vs NC-Net on PF-WILLOW at alpha=0.10NC-Net · PF-WILLOWPCK alpha=0.10 (%): 72.7 -> 73.8Improved

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