Research map / Image generation / GANs
Geometry Score: A Method For Comparing Generative Adversarial Networks
Read the paperCheck 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 |
|---|---|---|---|
| Use WGAN-GP instead of WGAN | MNIST GAN · MNIST all | Geometry Score (×10^3): 26.1 -> 2.04 | Helped |
| Use WGAN-GP instead of WGAN | MNIST GAN · MNIST classes 1, 2, 3, 5, 6, 7, 8 | Geometry Score (×10^3): 21.4 -> 1.44; 0.60 -> 0.54; 7.04 -> 0.27; 0.47 -> 0.03; 22.8 -> 13.5; 2.20 -> 1.38; 0.76 -> 0.14 | Helped |
| Use WGAN-GP instead of WGAN | MNIST GAN · MNIST classes 0, 4, 9 | Geometry Score (×10^3): 0.85 -> 5.19; 1.52 -> 2.16; 1.27 -> 5.00 | Hurt |
| Train DCGAN with the default, higher-capacity generator instead of a weak generator | CelebA GAN · CelebA | Geometry Score | Helped |
| Use Geometry Score to monitor training rather than relying on Inception Score alone | DCGAN on CelebA · CelebA | mode-collapse detection | Helped |
| Compare MRLT distributions with Geometry Score | CaloGAN · CaloGAN eplus layers | topological similarity | No clear effect |
Closest papers by meaning
- A Domain Agnostic Measure for Monitoring and Evaluating GANs (NeurIPS 2019 · similarity 0.92)
- Topology Distance: A Topology-Based Approach for Evaluating Generative Adversarial Networks (AAAI 2021 · similarity 0.92)
- Quantitatively Evaluating GANs With Divergences Proposed for Training (ICLR 2018 · similarity 0.91)
- SMaRt: Improving GANs with Score Matching Regularity (ICML 2024 · similarity 0.90)
- Are GANs Created Equal? A Large-Scale Study (NeurIPS 2018 · similarity 0.90)
- The GAN is dead; long live the GAN! A Modern GAN Baseline (NeurIPS 2024 · 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.