Research map / Image generation / GANs
GANs for Biological Image Synthesis
Anton Osokin, Anatole Chessel, Rafael E. Carazo Salas, Federico Vaggi
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 |
|---|---|---|---|
| train with WGAN-GP instead of GAN | separable DCGAN generator · LIN yeast cells C2ST (6 proteins) | C2ST score: 3.2 ± 0.1 -> 1.6 ± 0.1 | Improved |
| train with WGAN-GP instead of GAN | separable generator · LIN yeast cells regular reconstruction | L2-error: 0.053 ± 0.007 -> 0.043 ± 0.006 | Improved |
| use star-shaped generator instead of one-class separable generator | WGAN-GP · LIN yeast cells regular reconstruction | L2-error: 0.043 ± 0.006 -> 0.041 ± 0.005 | No clear effect |
| use star-shaped generator that synchronizes channels instead of multi-channel separable generator | WGAN-GP multi-channel generation · LIN yeast cells C2ST (6 proteins) | C2ST score: 3.8 ± 0.2 -> 1.4 ± 0.1 | Improved |
| synchronize green channels across classes with a star-shaped model instead of independent one-class generators | WGAN-GP · LIN yeast cells C2ST (6 proteins) | C2ST score: 0.8 ± 0.2 -> 1.4 ± 0.1 | Hurt |
| separate red and green channel filters (separable generator) in one-class generator | one-class WGAN-GP · LIN yeast cells C2ST (6 proteins) | C2ST score: 0.8 ± 0.2 -> 1.1 ± 0.2 | Hurt |
| use separable (red-first) reconstruction instead of regular reconstruction | WGAN-GP-star · LIN yeast cells reconstruction | L2-error: 0.041 ± 0.005 -> 0.058 ± 0.010 | Hurt |
| apply batch normalization in the red tower once instead of once per green tower (ablation) | star-shaped generator · LIN yeast cells | training convergence | Improved |
| use WGAN or WGAN-GP instead of cross-entropy classifier in C2ST (ablation) | C2ST evaluation · LIN yeast cells C2ST | C2ST variance | Improved |
Closest papers by meaning
- StudioGAN: A Taxonomy and Benchmark of GANs for Image Synthesis (TPAMI 2023 · similarity 0.86)
- Dynamically Grown Generative Adversarial Networks (AAAI 2021 · similarity 0.86)
- Mol2Image: Improved Conditional Flow Models for Molecule to Image Synthesis (CVPR 2021 · similarity 0.86)
- MSG-GAN: Multi-Scale Gradients for Generative Adversarial Networks (CVPR 2020 · similarity 0.86)
- Large Scale GAN Training for High Fidelity Natural Image Synthesis (ICLR 2019 · similarity 0.86)
- Unlocking Pre-trained Image Backbones for Semantic Image Synthesis (CVPR 2024 · similarity 0.85)
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.
