Research map / Image generation / GANs
Logo Synthesis and Manipulation With Clustered Generative Adversarial Networks
Alexander Sage, Eirikur Agustsson, Radu Timofte, Luc Van Gool
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 |
|---|---|---|---|
| Use auxiliary-classifier conditioning with 32 ResNet-feature clusters instead of unconditional iWGAN | iWGAN · CIFAR-10 | Inception score: 7.853$\pm$0.072 -> 8.673$\pm$0.075 | Improved |
| Use layer conditioning with 32 ResNet-feature clusters instead of unconditional iWGAN | iWGAN · CIFAR-10 | Inception score: 7.853$\pm$0.072 -> 7.831$\pm$0.072 | No clear effect |
| Use ResNet-feature clustering rather than autoencoder clustering with AC conditioning and 32 clusters | iWGAN-AC · CIFAR-10 | Inception score: 7.885$\pm$0.083 -> 8.673$\pm$0.075 | Improved |
| Use auxiliary-classifier conditioning rather than layer conditioning with 32 ResNet-feature clusters | iWGAN with RC clustering · CIFAR-10 | Inception score: 7.831$\pm$0.072 -> 8.673$\pm$0.075 | Improved |
| Increase ResNet-feature cluster count from 10 to 32 with AC conditioning (ablation) | iWGAN-AC · CIFAR-10 | Inception score: 8.433$\pm$0.068 -> 8.673$\pm$0.075 | Improved |
| Increase ResNet-feature cluster count from 32 to 128 with AC conditioning (ablation) | iWGAN-AC · CIFAR-10 | Inception score: 8.673$\pm$0.075 -> 8.625$\pm$0.109 | No clear effect |
| Increase ResNet-feature cluster count from 32 to 128 with AC conditioning (ablation) | iWGAN-AC · CIFAR-10 | MS-SSIM: 0.0500$\pm$0.0016 -> 0.0465$\pm$0.0015 | Improved |
| Use layer conditioning with 16 ResNet-feature clusters rather than unconditional iWGAN | iWGAN · LLD-icon-sharp | CORNIA score: 54.27$\pm$0.67 -> 55.37$\pm$0.67 | Hurt |
| Use layer conditioning with 128 ResNet-feature clusters rather than unconditional iWGAN | iWGAN · LLD-icon-sharp | CORNIA score: 54.27$\pm$0.67 -> 55.27$\pm$0.68 | Hurt |
| Use layer conditioning with 100 autoencoder clusters rather than DCGAN-LC | GAN on LLD-icon · LLD-icon | CORNIA score: 62.12$\pm$0.51 -> 60.24$\pm$0.61 | Improved |
| Add the latent-space and cluster-label directional shifts together rather than applying just one space's shift (ablation) | iWGAN-LC logo manipulation · LLD-logo | visual attribute shift quality | Improved |
Closest papers by meaning
- DeLiGAN : Generative Adversarial Networks for Diverse and Limited Data (CVPR 2017 · similarity 0.88)
- Large Scale GAN Training for High Fidelity Natural Image Synthesis (ICLR 2019 · similarity 0.86)
- Cluster-Guided Image Synthesis With Unconditional Models (CVPR 2022 · similarity 0.86)
- SketchyGAN: Towards Diverse and Realistic Sketch to Image Synthesis (CVPR 2018 · similarity 0.86)
- MSG-GAN: Multi-Scale Gradients for Generative Adversarial Networks (CVPR 2020 · similarity 0.86)
- StudioGAN: A Taxonomy and Benchmark of GANs for Image Synthesis (TPAMI 2023 · similarity 0.86)
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.
