AI Paper Atlas

Research map / Point clouds / Point cloud classification

Bootstrap Your Own Correspondences

Mohamed El Banani, Justin Johnson

ICCV 2021 · Point clouds › Point cloud classification · approach: Self-supervised point cloud pretraining · 34 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 visual bootstrapping (visual correspondences) to geometric training (ablation)BYOC · ScanNet registrationrotation accuracy at 5 degrees (%): 80.3 -> 86.5Improved
add visual bootstrapping (visual correspondences) (ablation)BYOC · 3D Match FMRfeature-match recall: 0.786 -> 0.766Hurt
add rotation augmentation (ablation)BYOC · 3D Match FMRfeature-match recall: 0.766 -> 0.827Improved
add rotation augmentation (ablation)BYOC · ScanNet registrationrotation accuracy at 5 degrees (%): 86.5 -> 71.2Hurt
train on ScanNet instead of 3D Match (ablation)BYOC · ScanNet registrationrotation accuracy at 5 degrees (%): 66.5 -> 86.5Improved
train on ScanNet instead of 3D Match (ablation)BYOC · 3D Match FMRfeature-match recall: 0.690 -> 0.766Improved
use Lowe-ratio filtered correspondences instead of all features (ablation)BYOC · 3D Match FMRfeature-match recall: 0.662 -> 0.766Improved
use Weighted Procrustes with Lowe ratio instead of RANSAC (ablation)BYOC · ScanNet registrationrotation accuracy at 5 degrees (%): 81.3 -> 86.5Improved
use RANSAC instead of Weighted Procrustes (ablation)FCGF · ScanNet registrationrotation accuracy at 5 degrees (%): 54.1 -> 75.3Improved
use visual features instead of geometric features for registration (ablation)BYOC · ScanNet registrationrotation mean error (degrees): 3.8 -> 2.7Improved
use random visual features instead of random geometric features (ablation)untrained encoders · ScanNet registrationrotation mean error (degrees): 21.3 -> 6.4Improved
replace pose-supervised FCGF+RANSAC by unsupervised BYOCregistration pipeline · ScanNet registrationrotation accuracy at 5 degrees (%): 75.3 -> 86.5Improved

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