AI Paper Atlas

Research map / Point clouds / Shape and surface completion

Discrete Point Flow Networks for Efficient Point Cloud Generation

Roman Klokov, Edmond Boyer, Jakob Verbeek

ECCV 2020 · Point clouds › Shape and surface completion · approach: Unsupervised shape correspondence · 103 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
replace continuous flows (PointFlow) with discrete affine coupling flows (DPF-Net)PointFlow-style model · ShapeNet generationtotal training time (days): 80 -> 1.1Improved
replace continuous flows with discrete affine coupling flowsPointFlow-style model · ShapeNet generationgeneration time (ms/sample): 150 -> 4Improved
replace continuous flows with discrete coupling flowsPointFlow-style model · ShapeNet generationtraining memory (Mb/sample): 470 -> 370Improved
train DPF-Net on normalized rather than non-normalized meshes (ablation)DPF-Net · ShapeNet autoencodingCD (x10^4): 6.85 -> 6.17Improved
train DPF-Net on normalized rather than non-normalized meshes (ablation)DPF-Net · ShapeNet autoencodingEMD (x10^2): 5.06 -> 4.37Improved
DPF-Net instead of PointFlow under equal training timeautoencoder · ShapeNet autoencodingCD (x10^4): 10.22 -> 6.85Improved
DPF-Net versus PointFlow on car classgenerative model · ShapeNet car generation1-NNA CD (%): 70.1 -> 70.3Hurt
DPF-Net versus PointFlow on airplanegenerative model · ShapeNet airplane generationJSD (x10^2): 1.42 -> 0.94Improved
DPF-Net versus l-GAN-CD on EMD-based 1-NNAgenerative model · ShapeNet airplane generation1-NNA EMD (%): 92.1 -> 67.0Improved
DPF-Net versus AtlasNet in single-view reconstructionsingle-view reconstruction · ShapeNet SVRCD (x10^3): 5.34 -> 5.51Mixed
DPF-Net versus voxel-based PRNsingle-view reconstruction · ShapeNet SVRF1 (%): 53.1 -> 52.4No clear effect

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