AI Paper Atlas

Research map / Label-efficient and robust learning / Recognition and long-tail

MVP-N: A Dataset and Benchmark for Real-World Multi-View Object Classification

Ren Wang, Jiayue Wang, Tae Sung Kim, JINSUNG KIM, Hyuk-Jae Lee

NeurIPS 2022 · Label-efficient and robust learning › Recognition and long-tail · approach: Uncertainty-aware multi-view learning · 5 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 max-pooling aggregation with DAN self-attentionMVCNN-new · MVP-N testMVA: 89.35 ± 1.21 -> 91.61 ± 0.94Improved
add GVCNN view groupingMVCNN-new · MVP-N testMVA: 89.35 ± 1.21 -> 85.42 ± 1.37Hurt
add CVR optimal-transport canonical views and transformerMVCNN-new · MVP-N testMVA: 89.35 ± 1.21 -> 79.99 ± 2.52Hurt
use predicted class probabilities as soft labels (SB vs HB)hard bootstrapping · MVP-N testMCDU: 0.3899 ± 0.0068 -> 0.2862 ± 0.0107Improved
use predicted class probabilities as soft labels (SB vs HB)hard bootstrapping · MVP-N testMVA: 83.20 ± 1.11 -> 83.85 ± 0.84Improved
train with HPIQ human information-quantity soft labels instead of CEcross-entropy · MVP-N testMVA: 83.37 ± 1.05 -> 94.36 ± 0.56Improved
train with HPIQ soft labels instead of CEcross-entropy · MVP-N testSVA: 78.65 ± 0.44 -> 63.31 ± 0.38Hurt
SAT instead of SEALSEAL · MVP-N testMVA: 86.42 ± 0.74 -> 87.37 ± 1.15Improved
label flip components (HB, DHB, LRT, PLC)CE · MVP-N testMVA: 83.37 ± 1.05 -> 82.96 ± 1.06No clear effect
increase uninformative views from 2-view to 6-view caseHPIQ · MVP-N testMVA: 97.73 ± 0.38 -> 92.07 ± 0.91Hurt
DAN instead of soft label methods at six viewssoft label methods · MVP-NMVAMixed

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