AI Paper Atlas

Research map / Point clouds / Point cloud classification

DH3D: Deep Hierarchical 3D Descriptors for Robust Large-Scale 6DoF Relocalization

Juan Du, Rui Wang, Daniel Cremers

ECCV 2020 · Point clouds › Point cloud classification · approach: Local feature aggregation networks · 117 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 self-supervision via synthesized transformations instead of weak submap-level supervision (ablation)DH3D local part · Oxford RobotCar registrationregistration performanceImproved
add the keypoint detector confidence map during training (ablation)DH3D local descriptor · Oxford RobotCar registrationregistration performanceImproved
add SE blocks to the local encoder (ablation)DH3D local descriptor · Oxford RobotCar registrationregistration performanceImproved
describe-and-detect rather than detect-then-describeDH3D detector · Oxford RobotCar repeatability, 256 keypointsrelative repeatability: roughly 1.75x than the best follower 3DFeatNetImproved
use DH3D learned detector instead of random keypointsDH3D descriptor · Oxford RobotCar registrationsuccess rate (%): 92.1 -> 98.5Improved
move from LiDAR point clouds to Stereo DSO point clouds without fine-tuning3DFeatNet + 3DFeatNet · Stereo DSO registrationsuccess rate (%): 98.1 -> 84.1Hurt
move from LiDAR point clouds to Stereo DSO point clouds without fine-tuningUSIP + 3DSmoothNet · Stereo DSO registrationsuccess rate (%): 98.0 -> 77.3Hurt
move from LiDAR point clouds to Stereo DSO point clouds without fine-tuningUSIP + 3DFeatNet · Stereo DSO registrationsuccess rate (%): 99.1 -> 69.9Hurt
move from LiDAR point clouds to Stereo DSO point clouds without fine-tuningDH3D detector + DH3D descriptor · Stereo DSO registrationsuccess rate (%): 98.5 -> 90.6Hurt
halve the points entering NetVLAD (4096 instead of 8192)DH3D global descriptor · Oxford RobotCar retrievalaverage recallNo clear effect
add Gaussian noise above 0.2mDH3D registration · Oxford RobotCar registrationsuccess rate: 0.15Hurt

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