AI Paper Atlas

Research map / 3D vision / 3D generation

Volume Transformer: Revisiting Vanilla Transformers for 3D Scene Understanding

Kadir Yilmaz, Adrian Kruse, Tristan Höfer, Daan de Geus, Bastain Leibe

ECCV 2026 · 3D vision › 3D generation · approach: Semantic scene completion · 1 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 data-efficient recipe (augmentation, regularization, CNN distillation) (ablation)Volt · ScanNet200 valmIoU: 31.0 -> 40.0Improved
add the same recipe upgrades to PTv3 (ablation)PTv3 · ScanNet200 valmIoU: 35.2 -> 37.5Improved
add joint multi-dataset trainingVolt-S · ScanNet valmIoU: 77.2 -> 80.2Improved
scale backbone from Volt-S to Volt-B under joint trainingVolt · ScanNet200 valmIoU: 38.1 -> 40.0Improved
use asymmetric instead of symmetric RoPE allocation (ablation)Volt-S · ScanNet200 / SemanticKITTImIoU: 35.4 -> 36.2; 69.5 -> 70.5Improved
keep metric token indices instead of per-scene normalization (ablation)Volt-S · ScanNet200 / SemanticKITTImIoU: 35.7 -> 36.2; 69.3 -> 70.5Improved
use RoPE instead of Fourier positional encoding (ablation)Volt-S · ScanNet200 / SemanticKITTImIoU: 28.6 -> 36.2; 66.0 -> 70.5Improved
add lightweight decoder (ablation)Volt-S · ScanNet200 / SemanticKITTImIoU: 35.0 -> 36.2; 69.9 -> 70.5Improved
use larger decoder with two residual conv blocks (ablation)Volt-S · ScanNet200 / SemanticKITTImIoU: 36.2 -> 35.8; 70.5 -> 69.5Hurt
use smaller 3x3x3 patches instead of 5x5x5 (ablation)Volt-S · ScanNet200 / SemanticKITTImIoU: 36.2 -> 37.7; 70.5 -> 71.0 (27 ms -> 67 ms)Mixed
use larger 7x7x7 patches instead of 5x5x5 (ablation)Volt-S · ScanNet200 / SemanticKITTImIoU: 36.2 -> 33.5; 70.5 -> 69.6Hurt
replace MinkUNet backbone with Volt-S in SPFormerSPFormer · ScanNet200 valmAP50: 33.8 -> 48.4Improved

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