AI Paper Atlas

Research map / Video / Video segmentation and tracking

SVQNet: Sparse Voxel-Adjacent Query Network for 4D Spatio-Temporal LiDAR Semantic Segmentation

Xuechao Chen, Shuangjie Xu, Xiaoyi Zou, Tongyi Cao, Dit-Yan Yeung, Lu Fang

ICCV 2023 · Video › Video segmentation and tracking · approach: Video semantic and panoptic segmentation · 12 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 Context Activator to a naive frame-stacking backbone (ablation)DRINet multi-scan backbone · SemanticKITTI val seq 08mIoU (%): 52.8 -> 53.9Improved
add Sparse Voxel-Adjacent Query module (ablation)backbone with CA · SemanticKITTI val seq 08mIoU (%): 53.9 -> 54.8Improved
add Temporal Feature Inheritance (ablation)backbone with CA and SVAQ · SemanticKITTI val seq 08mIoU (%): 54.8 -> 55.3Improved
add Context Activator to a stacking backbone (ablation)DRINet multi-scan backbone · SemanticKITTI val seq 08latency: 125 ms -> 61 msImproved
keep 45k rather than 40k activated historical voxels (ablation)Context Activator · SemanticKITTI val seq 08mIoU (%): 55.3 -> 55.2Hurt
reduce activated historical voxels from 40k to 30k (ablation)Context Activator · SemanticKITTI val seq 08mIoU (%): 55.3 -> 54.9Hurt
replace frame stacking with SVQNet modules on the same inputDRINet backbone · SemanticKITTI val seq 08, beyond 36mmIoU (%): 36.1 -> 42.3Improved
plug CA, SVAQ and TFI into SPVCNN instead of stackingSPVCNN · SemanticKITTI val multi-scanmIoU (%): 50.04 -> 52.36Improved
use SVQNet instead of Backbone+ stackingDRINet backbone · SemanticKITTI val seq 08GPU memory (MB): 8893 -> 6854Improved

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