Research map / Label-efficient and robust learning / Self-supervised and contrastive
Self-Supervised GANs via Auxiliary Rotation Loss
Check your idea against 108,967 papers
AI-extracted summary. Written by AI from the paper's title only (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 rotation-based self-supervision | Uncond-GAN · CIFAR10 | FID: 19.73 -> 17.11 | Helped |
| add rotation-based self-supervision | Uncond-GAN · ImageNet | FID: 56.67 -> 47.56 | Helped |
| add rotation-based self-supervision | Uncond-GAN · LSUN-bedroom | FID: 16.02 -> 13.66 | Helped |
| add rotation-based self-supervision | Uncond-GAN · CelebA-HQ | FID: 23.77 -> 26.11 | Hurt |
| add self-modulated batch normalization (ablation) | SS-GAN · CIFAR10 | FID: 17.11 -> 15.65 | Helped |
| add self-modulated batch normalization (ablation) | SS-GAN · ImageNet | FID: 47.56 -> 43.87 | Helped |
| add self-modulated batch normalization (ablation) | SS-GAN · LSUN-bedroom | FID: 13.66 -> 13.30 | Helped |
| add self-modulated batch normalization (ablation) | SS-GAN · CelebA-HQ | FID: 26.11 -> 24.36 | Helped |
| add rotation-based self-supervision | large-scale Uncond-GAN · ImageNet (50k evaluation) | FID: 71.6 ± 66.3 -> 23.6 ± 0.1 | Helped |
| remove adversarial loss, retaining rotation loss only (ablation) | SS-GAN · CIFAR10 | top-1 accuracy: 0.799 -> 0.762 | Hurt |
| remove adversarial loss, retaining rotation loss only (ablation) | SS-GAN · ImageNet | top-1 accuracy: 0.383 -> 0.202 | Hurt |
| add rotation-based self-supervision and self-modulated batch normalization | Uncond-GAN · ImageNet | top-1 accuracy: 0.083 -> 0.383 | Helped |
Closest papers by meaning
- On Self-Supervised Image Representations for GAN Evaluation (ICLR 2021 · similarity 0.93)
- Transformation GAN for Unsupervised Image Synthesis and Representation Learning (CVPR 2020 · similarity 0.91)
- Self-supervised GAN: Analysis and Improvement with Multi-class Minimax Game (NeurIPS 2019 · similarity 0.91)
- Improved Techniques for Training GANs (NeurIPS 2016 · similarity 0.90)
- Taming the Tail in Class-Conditional GANs: Knowledge Sharing via Unconditional Training at Lower Resolutions (CVPR 2024 · similarity 0.90)
- SAN: Inducing Metrizability of GAN with Discriminative Normalized Linear Layer (ICLR 2024 · similarity 0.88)
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.