AI Paper Atlas

Research map / Label-efficient and robust learning / Self-supervised and contrastive

Towards Comprehensive Representation Enhancement in Semantics-Guided Self-Supervised Monocular Depth Estimation

Jingyuan Ma, Xiangyu Lei, Nan Liu, Xian Zhao, Shiliang Pu

ECCV 2022 · Label-efficient and robust learning › Self-supervised and contrastive · approach: Pretext-task visual pre-training · 26 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 full IC-MHA (intra + cross attention) and LNBT over the CMA + LBT baseline (ablation)FSRE baseline · KITTI Eigen (R18)SqRel: 0.730 -> 0.690Improved
add full IC-MHA and LNBT over baseline (ablation)FSRE baseline · KITTI Eigen (R18)AbsRel: 0.107 -> 0.104Improved
replace CMA with intra-task-only attention IC-MHA* (ablation)FSRE baseline · KITTI Eigen (R18)AbsRel: 0.107 -> 0.106Improved
replace CMA with IC-MHA(All) (intra + cross) (ablation)FSRE baseline · KITTI Eigen (R18)AbsRel: 0.107 -> 0.105Improved
replace CMA with IC-MHA(All) (ablation)FSRE baseline · KITTI Eigen (R18)SqRel: 0.730 -> 0.734No clear effect
add LBT and LCE-based semantic supervision with IC-MHA vs photometric only (ablation)IC-MHA · KITTI Eigen (R18)AbsRel: 0.110 -> 0.105Improved
add LNBT to IC-MHA with LCE and LBT (ablation)IC-MHA · KITTI Eigen (R18)SqRel: 0.734 -> 0.690Improved
add LNBT alone instead of LBT (ablation)IC-MHA · KITTI Eigen (R18)SqRel: 0.794 -> 0.727Improved
add LBT and LNBT together (ablation)IC-MHA · KITTI Eigen (R18)AbsRel: 0.105 -> 0.104Improved
use proposed IC-MHA for joint depth-segmentationFSRE · KITTI 2015 segmentationmIoU: 55.8 -> 56.3Improved
switch from ResNet-18 to ResNet-50 at 192x640Ours · KITTI EigenSqRel: 0.690 -> 0.656Improved
raise resolution to 320x1024 with ResNet-50Ours · KITTI EigenAbsRel: 0.102 -> 0.099Improved

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