AI Paper Atlas

Research map / 3D vision / Depth and stereo

Towards Better Generalization: Joint Depth-Pose Learning Without PoseNet

Wang Zhao, Shaohui Liu, Yezhi Shu, Yong-Jin Liu

CVPR 2020 · 3D vision › Depth and stereo · approach: Dense depth prediction and completion · 183 citations (OpenAlex)

Check 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
Use optical-flow correspondences and two-view geometry instead of a PoseNet pose regressor, with explicit depth-scale alignment by triangulationJoint depth-pose learning · NYUv2 depth testrel: 0.283 -> 0.201Improved
Use optical-flow pose recovery and triangulation-based depth-scale alignmentJoint depth-pose learning · KITTI Odometry Seq. 09, stride 3translation error: 41.24 -> 7.21Improved
Jointly train depth-pose learning with the optical-flow module rather than training FlowNet-onlyOurs optical-flow estimator · KITTI 2015 training setAll EPE: 8.97 -> 5.72Improved
Add reprojection loss Lp (ablation)Ours depth model · KITTI Eigen test splitAbsRel: 0.135 -> 0.130Improved
Add ImageNet pretraining (ablation)Ours depth model · KITTI Eigen test splitAbsRel: 0.130 -> 0.113Improved
Remove triangulation supervision loss Ld (ablation)Ours depth model · KITTI Eigen test splittraining convergenceHurt
Replace optical flow and two-view triangulation with a PoseNet-like pose branch (ablation)Joint depth-pose learning · NYUv2 depth testrel: 0.201 -> 0.283Hurt
Predict pose with PoseNet using optical flow as input (ablation)Joint depth-pose learning · NYUv2 depth testrel: 0.201 -> 0.221Hurt
Increase NYUv2 evaluation resolution to 448×576 (ablation)Ours depth model · NYUv2 depth testrel: 0.201 -> 0.189Improved
Use the proposed method instead of PoseNetTUM-RGBD visual odometry · TUM-RGBD fr3/cabinetabsolute translational RMSE (meter): 1.45 -> 1.09Improved
Use the proposed method instead of ORB-SLAM2TUM-RGBD visual odometry · TUM-RGBD fr2/deskabsolute translational RMSE (meter): 0.006 -> 0.52Hurt
Train on TUM-RGBD and transfer to NYUv2 rather than use the PoseNet baselineDepth model transfer · NYUv2AbsRel: 0.324 -> 0.276Improved

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