AI Paper Atlas

Research map / Label-efficient and robust learning / Self-supervised and contrastive

Discovering Visual Patterns in Art Collections With Spatially-Consistent Feature Learning

Xi Shen, Alexei A. Efros, Mathieu Aubry

CVPR 2019 · Label-efficient and robust learning › Self-supervised and contrastive · approach: Pretext-task visual pre-training · 100 citations (OpenAlex)

Check 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
Fine-tune ImageNet features using spatially consistent self-supervised matches on BrueghelResNet-18 features · BrueghelmAP (cosine similarity): 58.0 -> 75.3Improved
Apply the discovery score instead of cosine similarityOurs (trained on Brueghel) · BrueghelmAP: 75.3 -> 76.4Improved
Train the feature on Brueghel rather than LTLLResNet-18 features · BrueghelmAP (cosine similarity): 65.2 -> 75.3Improved
Train the feature on LTLL rather than ImageNet-only featuresResNet-18 features · BrueghelmAP (cosine similarity): 58.0 -> 65.2Improved
Fine-tune features on LTLL and use discoveryResNet-18 + discovery · LTLLclassification accuracy (%): 80.9 -> 88.5Improved
Use the discovery score instead of image-level max-pooled featuresResNet-18 · LTLLclassification accuracy (%): 59.8 -> 80.9Improved
Fine-tune features on Oxford5KResNet-18 + discovery · Oxford5Kretrieval mAP (%): 85.0 -> 85.7Improved
Train features on LTLL rather than Oxford5KResNet-18 + discovery · Oxford5Kretrieval mAP (%): 85.7 -> 83.6Hurt
Use Context Prediction training instead of spatial-consistency feature learningResNet-18 features · BrueghelmAP (cosine similarity): 58.0 -> 58.8No clear effect
Use a positive region farther from the proposal region (P12 or P14) instead of a nearby positive region (ablation)Spatially-consistent feature learning · Brueghelone-shot matching mAPImproved
Use positive regions close to the proposal region (ablation)Spatially-consistent feature learning · Brueghelone-shot matching mAPHurt

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