Research map / Image generation / GANs
GAN Compression: Efficient Architectures for Interactive Conditional GANs
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 |
|---|---|---|---|
| Compress CycleGAN with GAN Compression | CycleGAN · horse→zebra | FID: 61.53 -> 71.81 | Hurt |
| Use GAN Compression instead of Shu et al.'s method | CycleGAN · horse→zebra | FID: 96.15 -> 71.81 | Helped |
| Compress MUNIT with GAN Compression | MUNIT · edges→shoes | FID: 30.13 -> 31.48 | Hurt |
| Compress Pix2pix with GAN Compression | Pix2pix · edges→shoes | FID: 24.18 -> 26.60 | Hurt |
| Compress Pix2pix with GAN Compression | Pix2pix · Cityscapes | mIoU: 42.06 -> 40.77 | Hurt |
| Compress GauGAN with GAN Compression | GauGAN · Cityscapes | mIoU: 62.18 -> 61.22 | Hurt |
| Compress GauGAN with Fast GAN Compression | GauGAN · COCO-Stuff | mIoU: 38.38 -> 35.34 | Hurt |
| Compress GauGAN with Fast GAN Compression | GauGAN · COCO-Stuff | FID: 21.96 -> 25.06 | Hurt |
| Use intermediate feature distillation and the full compression method instead of from-scratch training (ablation) | Pix2pix · Cityscapes | mIoU: 32.57 -> 40.77 | Helped |
| Use intermediate distillation rather than teacher-output distillation (ablation) | Pix2pix · Cityscapes | mIoU: 35.67 -> 38.26 | Helped |
| Decompose ResBlock convolutions rather than downsample convolutions at ngf=64 (ablation) | CycleGAN generator · horse→zebra | FID: 68.72 -> 62.95 | Helped |
| Decompose ResBlock convolutions rather than upsample convolutions at ngf=16 (ablation) | CycleGAN generator · horse→zebra | FID: 95.54 -> 79.49 | Helped |
Closest papers by meaning
- Content-Aware GAN Compression (CVPR 2021 · similarity 0.93)
- Revisiting Discriminator in GAN Compression: A Generator-discriminator Cooperative Compression Scheme (NeurIPS 2021 · similarity 0.91)
- Compressing Image-to-Image Translation GANs Using Local Density Structures on Their Learned Manifold (AAAI 2024 · similarity 0.91)
- AutoGAN-Distiller: Searching to Compress Generative Adversarial Networks (ICML 2020 · similarity 0.91)
- Towards Faster and Stabilized GAN Training for High-fidelity Few-shot Image Synthesis (ICLR 2021 · similarity 0.91)
- The GAN is dead; long live the GAN! A Modern GAN Baseline (NeurIPS 2024 · similarity 0.89)
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.