Research map / Image generation / Editing and style transfer
Content-Style Identification via Differential Independence
Subash Timilsina, Hoang-Son Nguyen, Sagar Shrestha, Xiao Fu
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 Jacobian orthogonality regularizer Lorth (ablation) | CSDI-GAN · AFHQ generation | FID: 5.3 -> 4.4 | Improved |
| add Jacobian orthogonality regularizer Lorth (ablation) | CSDI-GAN · CelebA-HQ generation | FID: 6.0 -> 4.3 | Improved |
| add Jacobian orthogonality regularizer Lorth (ablation) | CSDI-GAN · CelebA-HQ generation | LPIPS: 0.2467 -> 0.3392 | Improved |
| add Jacobian orthogonality regularizer Lorth (ablation) | CSDI-GAN · AFHQ generation | LPIPS: 0.4079 -> 0.4452 | Improved |
| add Lorth (ablation) | CSDI-GAN · two-domain MNIST generation | FID: 27.45 -> 23.80 | Improved |
| add Lorth (ablation) | CSDI-GAN · two-domain MNIST disentanglement | R2 style: 0.5618 -> 0.8995 | Improved |
| model dependent content and style instead of block independence (ablation) | CSDI-GAN w/o Lorth · CelebA-HQ generation | FID: 4.6 -> 6.0 | Hurt |
| replace B.I. GAN with CSDI-GAN | translation · AFHQ | FID: 10.5 -> 7.1 | Improved |
| replace B.I. GAN with CSDI-GAN | translation · CelebA-HQ | FID: 24.6 -> 12.9 | Improved |
| replace StarGANv2 with CSDI-GAN | translation · CelebA-HQ | LPIPS: 0.3148 -> 0.3105 | No clear effect |
| set lambda_inv or lambda_orth to zero (ablation) | CSDI-GAN · two-domain MNIST | FID | Hurt |
| use very large regularization weights (ablation) | CSDI-GAN · two-domain MNIST | FID | Hurt |
Earlier methods this paper compares with
Named baselines from earlier accepted papers in this paper's results tables: better than 1 of 1 on most shared results.
- StarGANv2 — StarGAN v2: Diverse Image Synthesis for Multiple Domains (CVPR 2020) · better on 3 of 4 shared results
Matched by method name in results tables read by AI. Authors' own tables tend to favour the newer method, and a comparison can be missed when a method is renamed.
Closest papers by meaning
- Content-Style Learning from Unaligned Domains: Identifiability under Unknown Latent Dimensions (ICLR 2025 · similarity 0.92)
- Counterfactual Generation with Identifiability Guarantees (NeurIPS 2023 · similarity 0.89)
- Unsupervised Disentanglement of Content and Style via Variance-Invariance Constraints (ICLR 2025 · similarity 0.86)
- Beyond ICA: Identifiability by Symmetry Breaking (NeurIPS 2026 · similarity 0.85)
- Conditional Balance: Improving Multi-Conditioning Trade-Offs in Image Generation (CVPR 2025 · similarity 0.85)
- Towards Identifiable Unsupervised Domain Translation: A Diversified Distribution Matching Approach (ICLR 2024 · similarity 0.84)
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.
