AI Paper Atlas

Research map / Point clouds / Point cloud classification

PointContrast: Unsupervised Pre-training for 3D Point Cloud Understanding

Saining Xie, Jiatao Gu, Demi Guo, Charles R. Qi, Leonidas Guibas, Or Litany

ECCV 2020 · Point clouds › Point cloud classification · approach: Self-supervised point cloud pretraining · 706 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 PointContrast unsupervised pre-training on ScanNet pairsSR-UNet · S3DIS Area 5 segmentationmIoU: 68.2 -> 70.9Improved
add PointContrast pre-trainingSR-UNet · ScanNet segmentationmIoU: 72.2 -> 74.1Improved
add PointContrast pre-trainingVoteNet with SR-UNet · ScanNet detectionmAP@0.5: 35.4 -> 38.0Improved
add PointContrast pre-trainingVoteNet with SR-UNet · SUN RGB-D detectionmAP@0.5: 31.7 -> 34.8Improved
add PointContrast pre-trainingSR-UNet · Synthia4D segmentationmIoU: 79.8 -> 83.1Improved
add PointContrast pre-training with 1% labeled dataSR-UNet · ShapeNet classificationaccuracy: 62.2 -> 66.2Improved
add PointContrast pre-training with all training dataSR-UNet · ShapeNet classificationaccuracy: 85.1 -> 85.7Improved
add PointContrast pre-training with 1% labeled dataSR-UNet · ShapeNet part segmentationIoU: 71.8 -> 74.0Improved
add ShapeNet supervised pre-training before fine-tuningSR-UNet · S3DIS segmentationtraining curveHurt
use PointInfoNCE instead of hardest-contrastive loss (ablation)PointContrast · Synthia4D segmentationmIoU: 82.6 -> 83.1Improved
use PointInfoNCE instead of hardest-contrastive loss (ablation)PointContrast · S3DIS Area 5 segmentationmIoU: 70.9 -> 70.3Hurt
use multi-view partial scans instead of whole scene as a single view (ablation)PointContrast · S3DIS Area 5 segmentationmIoU: 68.17 -> 68.35Improved

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