Research map / Architectures and efficiency / CNN design
Shift-Net: Image Inpainting via Deep Feature Rearrangement
Zhaoyi Yan, Xiaoming Li, Mu Li, Wangmeng Zuo, Shiguang Shan
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 |
|---|---|---|---|
| shift the connection at (L-4)-th layer instead of (L-3)-th (ablation) | Shift-Net · 256x256 image inference | runtime per image: 80 ms -> 40 ms | Mixed |
| shift the connection at (L-2)-th layer instead of (L-3)-th (ablation) | Shift-Net · 256x256 image inference | runtime per image: 80 ms -> 400 ms | Mixed |
| add guidance loss (ablation) | U-Net and Shift-Net · Paris StreetView | visual quality (qualitative) | Improved |
| use guidance weight 0.001 or at least 0.1 instead of 0.01 (ablation) | Shift-Net · Paris StreetView | visual quality (qualitative) | Hurt |
| add shift operation to U-Net (with guidance loss) (ablation) | U-Net · Paris StreetView | visual quality (qualitative) | Improved |
| use nearest neighbor search instead of random shift-connection (ablation) | Shift-Net · Paris StreetView | visual quality (qualitative) | Improved |
| vary mask threshold T among 4/16, 5/16, 6/16 (ablation) | Shift-Net · Paris StreetView | visual quality (qualitative) | No clear effect |
| combine CNN with exemplar-based inpainting end-to-end (Shift-Net vs MNPS) | Shift-Net · Paris StreetView | PSNR: 25.98 -> 26.51 | Improved |
Later papers that compared with this method
1 later accepted paper reports Shift-Net-18-9 (1.76M) as a baseline in their results tables; 1 of them report a better number on most of the results they share with it.
- Super Sparse Convolutional Neural Networks (AAAI 2019 · better on 6 of 6 shared results)
Matched by method name in results tables read by AI. Authors' own tables tend to favour the newer method, and a comparison can be missed when a method is renamed.
Closest papers by meaning
- Learning Pyramid-Context Encoder Network for High-Quality Image Inpainting (CVPR 2019 · similarity 0.89)
- Contextual-based Image Inpainting: Infer, Match, and Translate (ECCV 2018 · similarity 0.89)
- Coherent Semantic Attention for Image Inpainting (ICCV 2019 · similarity 0.88)
- Shift: A Zero FLOP, Zero Parameter Alternative to Spatial Convolutions (CVPR 2018 · similarity 0.88)
- Generative Image Inpainting With Contextual Attention (CVPR 2018 · similarity 0.88)
- SyFormer: Structure-Guided Synergism Transformer for Large-Portion Image Inpainting (AAAI 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.
