AI Paper Atlas

Research map / Label-efficient and robust learning / Domain adaptation and generalization

Few-shot Image Generation with Elastic Weight Consolidation

Yijun Li, Richard Zhang, Jingwan (Cynthia) Lu, Eli Shechtman

NeurIPS 2020 · Label-efficient and robust learning › Domain adaptation and generalization · approach: Cross-domain few-shot with transformers

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
Use Fisher-weighted EWC while adapting the pretrained generatorStyleGAN few-shot image generation · 10-shot target domainsFID: 86.44~$\pm$~4.38 -> 74.87~$\pm$~3.75Improved
Use Fisher-weighted EWC while adapting the pretrained generatorStyleGAN few-shot image generation · 10-shot target domainsLPIPS diversity: 0.21~$\pm$~0.02 -> 0.40~$\pm$~0.02Improved
Use Fisher-weighted EWC while adapting the pretrained generatorStyleGAN few-shot image generation · 10-shot target domainsUser fooling rate (%): 30.56~$\pm$~7.14 -> 47.92~$\pm$~7.38Improved
Use EWC regularization instead of setting λ=0 (ablation)StyleGAN adapted to the target domain · 10-shot Fernand Léger facesoverfitting to training examplesImproved
Remove EWC regularization by setting λ=0 (ablation)StyleGAN adapted to the target domain · 10-shot Fernand Léger facesdiversity of generated resultsHurt
Increase the number of target examplesOurs (EWC) · 10-shot to 10,000-shot target domainsFID: 74.87~$\pm$~3.75 -> 55.74~$\pm$~2.88Improved
Increase the number of target examplesBSA · 10-shot to 10,000-shot target domainsFID: 105.56~$\pm$~5.79 -> 131.20~$\pm$~6.04Hurt
Increase the number of target examplesNST · 1-shot to 10,000-shot target domainsFID: 212.23~$\pm$~9.77 -> 194.88~$\pm$~8.71No clear effect
Increase the number of target examplesMineGAN · 1-shot to 10,000-shot target domainsFID: 102.57~$\pm$~4.76 -> 58.69~$\pm$~3.14Improved
Choose a source domain more similar to the targetPretrained generator adaptation · 10-shot cross-domain generationFIDImproved
Use a larger regularization weight when source and target are more similarEWC adaptation · Cross-domain generationadaptation performanceImproved
Use a smaller regularization weight when more target data is availableEWC adaptation · Cross-domain generationadaptation performanceImproved

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