AI Paper Atlas

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

ALADIN: All Layer Adaptive Instance Normalization for Fine-Grained Style Similarity

Dan Ruta, Saeid Motiian, Baldo Faieta, Zhe Lin, Hailin Jin, Alex Filipkowski, Andrew Gilbert, John Collomosse

ICCV 2021 · Label-efficient and robust learning › Recognition and long-tail · approach: Correspondence and causal representation learning · 41 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 contrastive loss instead of triplet loss (ablation)ALADIN-L · BAM-FGIR Top-1: 26.70 -> 56.89Improved
add contrastive loss instead of listwise loss (ablation)ALADIN-L · BAM-FGIR Top-1: 37.90 -> 56.89Improved
add weak group supervision instead of reconstruction only (ablation)ALADIN-L · BAM-FGIR Top-1: 11.48 -> 56.89Improved
use ALADIN architecture instead of MUNIT style code under contrastive trainingcontrastive training · BAM-FGIR Top-1: 6.43 -> 56.89Improved
use larger ALADIN-L backbone instead of ALADIN-S (ablation)ALADIN contrastive · BAM-FGIR Top-1: 36.42 -> 56.89Improved
add projection MLP to listwise loss (ablation)ALADIN-L listwise · BAM-FGIR Top-1: 25.14 -> 37.90Improved
add projection MLP to triplet loss (ablation)ALADIN-L triplet · BAM-FGIR Top-1: 26.70 -> 22.12Hurt
add hard negative mining to triplet loss (ablation)ALADIN-L triplet · BAM-FGIR Top-1: 26.70 -> 23.75Hurt
add group supervision instead of SimCLR-style augmentation (ablation)ALADIN-L contrastive · BAM-FGIR Top-1: 30.47 -> 56.89Improved
add group supervision instead of augmentation (ablation)ResNet50 contrastive · BAM-FGIR Top-1: 28.51 -> 40.14Improved
increase training data from 250K to 1M images (ablation)ALADIN-L contrastive · BAM-FGIR Top-1: 40.08 -> 56.89Improved
fine-tune on cleaned BAM-FG-C5 data (ablation)ALADIN-L · BAM-FGIR Top-1: 56.89 -> 58.98Improved

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