AI Paper Atlas

Research map / 3D vision / Depth and stereo

Fine-Grained Semantics-Aware Representation Enhancement for Self-Supervised Monocular Depth Estimation

Hyunyoung Jung, Eunhyeok Park, Sungjoo Yoo

ICCV 2021 · 3D vision › Depth and stereo · approach: Self-supervised monocular depth · 140 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 segmentation loss L_CE via shared encoder (ablation)Monodepth2 · KITTI Eigen splitAbsRel: 0.114 -> 0.112Improved
add semantics-guided triplet loss (ablation)Monodepth2 + L_CE · KITTI Eigen splitAbsRel: 0.112 -> 0.108Improved
add CMA module (ablation)Monodepth2 + L_CE · KITTI Eigen splitAbsRel: 0.112 -> 0.107Improved
combine triplet loss and CMA module (ablation)Monodepth2 + L_CE · KITTI Eigen splitAbsRel: 0.112 -> 0.105Improved
train with high-resolution images (ablation)Full model · KITTI Eigen splitAbsRel: 0.105 -> 0.102Improved
use patch size K=5 instead of K=3 (ablation)L_SGT · KITTI Eigen splitAbsRel: 0.109 -> 0.108Improved
use patch size K=5 instead of K=7 (ablation)L_SGT · KITTI Eigen splitSqRel: 0.776 -> 0.755Improved
apply L_SGT to layer l=4 additionally (ablation)L_SGT · KITTI Eigen splitSqRel: 0.755 -> 0.742Mixed
make CMA bidirectional (refine depth and segmentation) (ablation)CMA module · KITTI Eigen splitAbsRel: 0.108 -> 0.107Improved
make CMA bidirectional versus segmentation-only target (ablation)CMA module · KITTI Eigen splitAbsRel: 0.111 -> 0.107Improved
use four embeddings instead of one in CMA (ablation)CMA module · KITTI Eigen splitSqRel: 0.762 -> 0.741Improved
use four embeddings instead of two in CMA (ablation)CMA module · KITTI Eigen splitAbsRel: 0.107 -> 0.107No 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