AI Paper Atlas

Research map / Architectures and efficiency / Neural architecture search

PVNAS: 3D Neural Architecture Search with Point-Voxel Convolution

Zhijian Liu, Haotian Tang, Shengyu Zhao, Kevin Shao, Song Han

TPAMI 2021 · Architectures and efficiency › Neural architecture search · approach: Latency-aware architecture search · 53 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
use sparse SPVConv at fine voxel size 0.05 instead of coarse dense PVConv (ablation)outdoor model · SemanticKITTI large scenesmIoU: 39.0 -> 58.8Improved
use PVConv instead of SPVConv for small objects (ablation)PVCNN · ShapeNetlatency speedup: 2.4x measured speedupImproved
use PVConv instead of SPVConv on PartNet (ablation)PVCNN (0.5xC) · PartNetmIoU: 2.2% higher accuracy, 3.8x fasterImproved
use PVConv instead of SPVConv on indoor scenes (ablation)PVCNN · S3DISmIoU: 1.2-2.1% higher mIoUImproved
replace MLPs in PointNet with PVConvPointNet · ShapeNetmIoU: 83.7 -> 86.2Improved
replace MLPs in PointNet with PVConvPointNet · S3DISmIoU: 43.0 -> 56.1Improved
add sparse point-voxel convolution and NAS to MinkowskiNetSPVNAS-B · SemanticKITTImIoU: 63.1 -> 66.4Improved
add 3D-NAS to SPVCNN with 15G MAC constraint (ablation)SPVNAS-A · SemanticKITTImIoU: 65.3 -> 63.7 with 118.6G -> 15.0 MACsImproved
add 3D-NAS to SPVCNN (ablation)SPVNAS · nuScenesmIoU: 54.7 mIoU with 23.1 -> 10.6 GMACsImproved
add 3D-NAS to PVCNN (PVNAS-B vs PVCNN 1xC) (ablation)PVNAS-B · PartNetmIoU: 47.8 -> 48.0Improved
use PVCNN in F-PointNet instead of PointNet++F-PointNet · KITTI car moderateAP: 72.0 -> 72.1No clear effect
replace SECOND sparse encoder with SPVCNNSECOND · KITTI car moderateAP: 80.9 -> 81.8Improved

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