AI Paper Atlas

Research map / Point clouds / Registration and matching

PointRegGPT: Boosting 3D Point Cloud Registration using Generative Point-Cloud Pairs for Training

Suyi Chen, Hao Xu, Haipeng Li, Kunming Luo, Guanghui Liu, Chi-Wing Fu, Ping Tan, Shuaicheng Liu

ECCV 2024 · Point clouds › Registration and matching · approach: Learned feature-based registration · 8 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 diffusion-generated pairs to GeoTrans trainingGeoTransformer · 3DLoMatchRegistration Recall: 74.0 -> 77.2Improved
generated data versus rendered SIRA data for GeoTransGeoTransformer · 3DMatchRegistration Recall: 94.1 vs 93.3No clear effect
add generated data to PREDATORPREDATOR · 3DLoMatchRegistration Recall: 56.2 -> 63.0Improved
add generated data to GeoTransGeoTransformer · 3DLoMatch (RANSAC, 5000 samples)Registration Recall: 70.4 -> 74.8 (vs reproduced); 75.0 -> 74.8 (vs reported)Mixed
add generated data and evaluate zero-shot on outdoor ETHPREDATOR · ETHRegistration Recall: 74.7 -> 84.1Improved
add generated data to GeoTrans on ETHGeoTransformer · ETH (LGR)Feature Matching Recall: 59.9 -> 56.1Hurt
add depth generation (diffusion inpainting) (ablation)PointRegGPT pipeline · 3DLoMatchRegistration Recall: 73.9 -> 76.5Improved
use re-projection only (no generation, no correction) (ablation)GeoTransformer · 3DLoMatchRegistration Recall: 74.0 -> 72.1Hurt
add depth correction module (ablation)PointRegGPT pipeline · 3DLoMatchRegistration Recall: 73.5 -> 76.5Improved
condition on real depth maps instead of unconditional generation (ablation)PointRegGPT pipeline · 3DLoMatchRegistration Recall: 73.2 -> 76.5Improved
add depth augmentation module to depth correction (ablation)depth correction · Depth correction (tau_m = 0.99)False positives: 3.17 -> 2.27Improved
increase generated pairs from 0 to 160k (ablation)CoFiNet · 3DLoMatchRegistration Recall: 61.4 -> 67.5Improved

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