Research map / Architectures and efficiency / Attention and transformers
Bridging Global Context Interactions for High-Fidelity Image Completion
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 |
|---|---|---|---|
| Use TFill instead of ICT | Image completion · Places2 free-form masks 20-50% | LPIPS: 0.1045 -> 0.0918; 0.1563 -> 0.1328; 0.1974 -> 0.1796 | Helped |
| Use TFill instead of ICT | Image completion · Places2 free-form masks 20-50% | FID: 17.13 -> 15.28; 22.39 -> 19.99; 28.18 -> 25.88 | Helped |
| Use TFill instead of PIC | Image completion · CelebA-HQ and FFHQ center masks | LPIPS: 0.061 -> 0.048; 0.068 -> 0.053 | Helped |
| Add attention in G to the Traditional Conv baseline (ablation) | TFill architecture · CelebA-HQ center mask | FID: 6.29 -> 6.34 | Hurt |
| Add a restrictive CNN token embedding (ablation) | TFill architecture · CelebA-HQ center mask | FID: 6.34 -> 4.68 | Helped |
| Add a transformer encoder (ablation) | TFill architecture · CelebA-HQ center mask | FID: 4.68 -> 4.02 | Helped |
| Add masked weights to transformer attention (ablation) | TFill architecture · CelebA-HQ center mask | FID: 4.02 -> 3.92 | Helped |
| Add the refinement network (ablation) | TFill architecture · CelebA-HQ center mask | FID: 3.92 -> 3.86 | Helped |
| Add a restrictive CNN token embedding (ablation) | TFill architecture · FFHQ center mask | FID: 4.01 -> 3.87 | Helped |
| Add a transformer encoder (ablation) | TFill architecture · FFHQ center mask | FID: 3.87 -> 3.66 | Helped |
| Add masked weights to transformer attention (ablation) | TFill architecture · FFHQ center mask | FID: 3.66 -> 3.63 | Helped |
| Add the refinement network (ablation) | TFill architecture · FFHQ center mask | FID: 3.63 -> 3.50 | Helped |
Closest papers by meaning
- Context and Geometry Aware Voxel Transformer for Semantic Scene Completion (NeurIPS 2024 · similarity 0.85)
- Revisiting Continuity of Image Tokens for Cross-domain Few-shot Learning (ICML 2025 · similarity 0.85)
- MaGIC: Multi-modality Guided Image Completion (ICLR 2024 · similarity 0.85)
- Memory-Augmented Re-Completion for 3D Semantic Scene Completion (AAAI 2025 · similarity 0.84)
- Masked Completion via Structured Diffusion with White-Box Transformers (ICLR 2024 · similarity 0.84)
- Unleashing Network Potentials for Semantic Scene Completion (CVPR 2024 · similarity 0.84)
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.