AI Paper Atlas

Research map / Point clouds / Point cloud classification

MATE: Masked Autoencoders are Online 3D Test-Time Learners

M. Jehanzeb Mirza, Inkyu Shin, Wei Lin, Andreas Schriebl, Kunyang Sun, Jaesung Choe, Mateusz Kozinski, Horst Possegger, In So Kweon, Kuk-Jin Yoon, Horst Bischof

ICCV 2023 · Point clouds › Point cloud classification · approach: Efficient point sampling architectures · 22 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 masked-autoencoder test-time training (MATE-Online) on top of joint-trained networkPointMAE · ModelNet-40CTop-1 accuracy (%): 53.9 -> 71.3Improved
apply entropy-minimization TTT (TENT, SHOT) at batch size 1PointMAE · ModelNet-40CTop-1 accuracy (%): 53.9 -> 26.5 (TENT), 26.6 (SHOT)Hurt
use rotation prediction as the TTT auxiliary task instead of MAE reconstruction (ablation)PointMAE · ModelNet-40CTop-1 accuracy (%): 53.9 -> 53.0 (TTT-Rot)Hurt
use larger test-time batch size (128) for baselines (ablation)TENT, SHOT, T3A · ModelNet-40CTop-1 accuracy (%): TENT 26.5 -> 65.6, SHOT 26.6 -> 63.8Improved
increase masking ratio from 60% to 95% (ablation)MATE-Online · ModelNet-40CTop-1 accuracy (%): 71.5 -> 71.6No clear effect
increase masking ratio to 97.5% (ablation)MATE-Online · ModelNet-40CTop-1 accuracy (%): 71.6 -> 56.9Hurt
increase TTT batch size from 1 to 2 (ablation)MATE-Online · ModelNet-40CTop-1 accuracy (%): 43.1 -> 66.4Improved
increase TTT batch size from 8 to 48 (ablation)MATE-Online · ModelNet-40CTop-1 accuracy (%): 69.7 -> 71.3Improved
adapt only every 5th sample (stride 5) (ablation)MATE · ShapeNet-CTop-1 accuracy (%): about 3 percent-point drop, 20 FPSHurt
adapt only every 300th sample (stride 300) (ablation)MATE · ShapeNet-CTop-1 accuracy (%): ~3 percent-points above Source-Only at 62 FPSImproved
use MATE-Standard (reset per sample) vs MATE-Online (accumulate) (ablation)MATE · ModelNet-40CTop-1 accuracy (%): 64.3 -> 71.3Improved
apply MATE to combined corruptionsPointMAE · ModelNet-40 combined corruptionsTop-1 accuracy (%): 30.6 -> 53.2Improved

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