AI Paper Atlas

Research map / 3D vision / Depth and stereo

Seeing Depth Through Frequency and Motion: A Progressive Training Paradigm for Monocular Depth Estimation

Ke Li, Bolin Song, Hongbo Liu

CVPR 2026 · 3D vision › Depth and stereo · approach: Self-supervised monocular depth

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 FGDepth (Frequency-Guided Sampling) to the depth network under joint training (ablation)baseline depth network · KITTI Eigen splitSq Rel: 0.727 -> 0.712Improved
add PQNet as pose network under joint training (ablation)baseline · KITTI Eigen splitSq Rel: 0.727 -> 0.723Improved
combine FGDepth and PQNet under joint training (ablation)baseline · KITTI Eigen splitSq Rel: 0.727 -> 0.704Improved
use progressive three-stage training instead of single-stage joint training (ablation)FGDepth + PQNet · KITTI Eigen splitSq Rel: 0.704 -> 0.679Improved
alternate depth and pose updates every mini-batch instead of three-stage training (ablation)FGDepth + PQNet · KITTI Eigen splitAbs Rel: 0.098 (joint) -> 0.117Hurt
alternate depth and pose updates every epoch instead of joint training (ablation)FGDepth + PQNet · KITTI Eigen splitSq Rel: 0.704 (joint) -> 0.710No clear effect
add PQNet to FGDepth under joint training (ablation)pose estimation · KITTI Odometry Seq. 09ATE: 0.019 -> 0.018Improved
use three-stage training instead of joint training (ablation)FGDepth + PQNet · KITTI Odometry Seq. 09ATE: 0.018 -> 0.016Improved
use three-stage training instead of joint training (ablation)FGDepth + PQNet · KITTI Odometry Seq. 10ATE: 0.015 -> 0.014No clear effect
use FGDepth instead of MonoViTKITTI depth model · KITTI Eigen splitSq Rel: 0.708 -> 0.679Improved
use FGDepth instead of MonoViT in zero-shot transferKITTI-trained depth model · Make3D (zero-shot)Sq Rel: 2.758 -> 2.627Improved

Earlier methods this paper compares with

Named baselines from earlier accepted papers in this paper's results tables: better than 1 of 1 on most shared results.

Matched by method name in results tables read by AI. Authors' own tables tend to favour the newer method, and a comparison can be missed when a method is renamed.

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