Research map / Image generation / GANs
Large Scale Image Completion via Co-Modulated 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 co-modulation rather than DeepFillv2 (retrained) | Image completion · FFHQ, large-scale completion | P-IDS (%): 0.9 \pm 0.1 -> 16.6 \pm 0.3 | Helped |
| Use co-modulation rather than DeepFillv2 (retrained) | Image completion · Places2, large-scale completion | FID: 22.1 \pm 0.1 -> 7.9 \pm 0.0 | Helped |
| Use co-modulation rather than SPADE | Image-to-image translation · COCO-Stuff | P-IDS (%): 1.1 -> 4.5 | Helped |
| Use co-modulation rather than SPADE | Image-to-image translation · COCO-Stuff | FID: 22.6 -> 22.5 | No clear effect |
| Add a stochastic latent style representation to conditional modulation (ablation) | Co-Mod on FFHQ · FFHQ, masked ratios (0.0, 0.2) through (0.8, 1.0) | FID: 0.55 \pm 0.01 -> 0.54 \pm 0.01; 1.78 \pm 0.02 -> 1.69 \pm 0.01; 3.03 \pm 0.02 -> 2.88 \pm 0.03; 4.42 \pm 0.02 -> 4.13 \p | Helped |
| Replace co-modulation with a vanilla conditional generator (ablation) | Image completion on FFHQ · FFHQ, masked ratio (0.8, 1.0) | FID: 5.52 \pm 0.08 -> 286.53 \pm 0.93 | Hurt |
| Increase truncation to ψ = 3 (ablation) | Co-Mod image-to-image translation · Edges2Handbags | LPIPS: 0.143 -> 0.379 | Helped |
| Increase truncation to ψ = 3 (ablation) | Co-Mod image-to-image translation · Edges2Handbags | FID: 56.9 -> 71.1 | Hurt |
Closest papers by meaning
- Diverse Shape Completion via Style Modulated Generative Adversarial Networks (NeurIPS 2023 · similarity 0.89)
- Pluralistic Image Completion (CVPR 2019 · similarity 0.89)
- The Missing Data Encoder: Cross-Channel Image Completion with Hide-and-Seek Adversarial Network (AAAI 2020 · similarity 0.88)
- Generative Face Completion (CVPR 2017 · similarity 0.88)
- Unleashing Network Potentials for Semantic Scene Completion (CVPR 2024 · similarity 0.86)
- MaGIC: Multi-modality Guided Image Completion (ICLR 2024 · 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.