Research map / Label-efficient and robust learning / Domain adaptation and generalization
MineGAN++: Mining Generative Models for Efficient Knowledge Transfer to Limited Data Domains
Yaxing Wang, Abel González-García, Chenshen Wu, Luis Herranz, Fahad Shahbaz Khan, Shangling Jui, Jian Yang, Joost van de Weijer
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 changed | Where | Measured | Effect |
|---|---|---|---|
| add Sparse Subnetwork Selection (S3) to MineGAN (ablation) | MineGAN · FFHQ Women (100 images) | FID: 93.1 -> 82.1 | Improved |
| add Sparse Subnetwork Selection (S3) to MineGAN (ablation) | MineGAN · FFHQ Children (100 images) | FID: 93.4 -> 84.9 | Improved |
| add finetuning stage after mining (ablation) | MineGAN (w/o FT) · FFHQ Women (100 images) | FID: 102.3 -> 93.1 | Improved |
| add Sparse Subnetwork Selection (S3) (ablation) | MineGAN · FFHQ (25 images, SNGAN) | KMMD: 0.337 -> 0.329 | Improved |
| add Sparse Subnetwork Selection to TransferGAN (ablation) | TransferGAN · Animal face (Bear) | FID: 82.82 -> 77.84 | Improved |
| select whole filters instead of individual weights (ablation) | Sparse Subnetwork Selection · StyleGAN transfer | FID | Improved |
| use 6 miner layers instead of 2 (ablation) | MineGAN miner · AFHQ cat (100 training images) | FID: 51.6 -> 58.3 | Hurt |
| use permutation-based improved multi-GAN optimization (ablation) | MineGAN(multi) · Red vehicle (0.5:0.5) | FID: 59.4 -> 53.2 | Improved |
| fuse pretrained GANs into one conditional GAN before mining (ablation) | MineGAN(multi) · Winter church | FID: 75.4 -> 69.3 | Improved |
| fuse pretrained GANs into one conditional GAN before mining (ablation) | MineGAN(multi) · Red vehicle (0.5:0.5) | FID: 53.2 -> 55.8 | Hurt |
| add Sparse Subnetwork Selection to fused mining (ablation) | MineGAN(fusion) · Red vehicle (0.5:0.5) | FID: 55.8 -> 48.5 | Improved |
| initialize critic from a different pretrained GAN (Kitchen vs Church) (ablation) | MineGAN(multi) · LSUN Tower | FID: 62.4 -> 61.7 | No clear effect |
Closest papers by meaning
- MineGAN: Effective Knowledge Transfer From GANs to Target Domains With Few Images (CVPR 2020 · similarity 0.98)
- Efficient Conditional GAN Transfer With Knowledge Propagation Across Classes (CVPR 2021 · similarity 0.88)
- DeGAN: Data-Enriching GAN for Retrieving Representative Samples from a Trained Classifier (AAAI 2020 · similarity 0.88)
- Transferring GANs: generating images from limited data (ECCV 2018 · similarity 0.88)
- Taming the Tail in Class-Conditional GANs: Knowledge Sharing via Unconditional Training at Lower Resolutions (CVPR 2024 · similarity 0.87)
- MCGAN: Enhancing GAN Training with Regression-Based Generator Loss (AAAI 2025 · similarity 0.87)
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.
