AI Paper Atlas

Research map / Detection and segmentation / CNN features

Semantic Line Detection Using Mirror Attention and Comparative Ranking and Matching

Dongkwon Jin, Jun-Tae Lee, Chang-Su Kim

ECCV 2020 · Detection and segmentation › CNN features · approach: Local features and context in CNNs · 13 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 mirror attention (with flipped feature map) to D-Net (ablation)D-Net+NMS · SELAUC A/P/R: 92.48 -> 93.38 (AUC A); 83.37 -> 83.98 (AUC P); 84.76 -> 86.23 (AUC R)Improved
add feature mirroring (flipped map) to the attention module (ablation)D-Net+NMS · SELAUC A/P/R: +0.14, +2.65, +1.37 (AUC A, P, R)Improved
use mirror attention instead of spatial-channel attention (ablation)D-Net+NMS · SELAUC A/P/R: 92.85 -> 93.38 (AUC A); 81.74 -> 83.98 (AUC P)Improved
add non-mirrored attention to D-Net (ablation)D-Net+NMS · SELAUC P: 83.37 -> 81.33 (AUC P); 92.48 -> 93.24 (AUC A)Mixed
replace NMS with R-Net and M-Net (ablation)D-Net with mirror attention · SELAUC A/P/R: +1.16, +1.46, +0.93 (AUC A, P, R)Improved
replace NMS with R-Net and M-Net (ablation)D-Net · AVA landscape dominant parallel lines, K=1AUC A: 52.62 -> 56.31Improved
replace NMS with R-Net and M-Net (ablation)D-Net · AVA landscape dominant parallel lines, K=5AUC A: 43.30 -> 50.72Improved
replace SLNet detector with D-Net (both with NMS) (ablation)SLNet · AVA landscape dominant parallel lines, K=1AUC A: 46.83 -> 52.62Improved
use DRM instead of SLNet (ablation)SLNet · SEL HardAUC A/P/R: +7.09, +12.97, +7.01 (AUC A, P, R)Improved
use DRM as vanishing point detector instead of NeurVPSNeurVPS · AVA landscape VP detectionAA 10 deg: 65.9 -> 68.3 (AA 10); worse at small tolerancesMixed

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