Research map / Image generation / Image restoration
Image Inpainting With Learnable Bidirectional Attention Maps
Chaohao Xie, Shaohui Liu, Chao Li, Ming-Ming Cheng, Wangmeng Zuo, Xiao Liu, Shilei Wen, Errui Ding
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 |
|---|---|---|---|
| make mask convolution filters learnable (instead of fixed 1/16) (ablation) | LBAM · Paris StreetView ablation, mask ratio (0.4, 0.5] | PSNR (dB): 21.20 -> 22.62 | Improved |
| make mask convolution filters learnable (instead of fixed 1/16) (ablation) | LBAM · Paris StreetView ablation, mask ratio (0.1, 0.2] | PSNR (dB): 26.95 -> 28.73 | Improved |
| add reverse attention maps in the decoder (ablation) | LBAM · Paris StreetView ablation, mask ratio (0.1, 0.2] | PSNR (dB): 27.80 -> 28.73 | Improved |
| add reverse attention maps in the decoder (ablation) | LBAM · Paris StreetView ablation, mask ratio (0.4, 0.5] | PSNR (dB): 21.76 -> 22.62 | Improved |
| use asymmetric Gaussian-shaped activation instead of ReLU (ablation) | LBAM · Paris StreetView ablation, mask ratio (0.1, 0.2] | PSNR (dB): 27.62 -> 28.73 | Improved |
| use asymmetric Gaussian-shaped activation instead of ReLU (ablation) | LBAM · Paris StreetView ablation, mask ratio (0.4, 0.5] | PSNR (dB): 21.48 -> 22.62 | Improved |
| use asymmetric Gaussian-shaped activation instead of LeakyReLU (ablation) | LBAM · Paris StreetView ablation, mask ratio (0.4, 0.5] | PSNR (dB): 18.38 -> 22.62 | Improved |
| use asymmetric Gaussian-shaped activation instead of Sigmoid (ablation) | LBAM · Paris StreetView ablation, mask ratio (0.4, 0.5] | PSNR (dB): 20.32 -> 22.62 | Improved |
| use 4x4 mask updating filter instead of 3x3 (ablation) | LBAM · Paris StreetView ablation, mask ratio (0.1, 0.2] | PSNR (dB): 28.74 -> 28.73 | No clear effect |
| add adversarial loss (ablation) | LBAM · Paris StreetView ablation, mask ratio (0.1, 0.2] | PSNR (dB): 29.19 -> 28.73 | Hurt |
| add adversarial loss (ablation) | LBAM · Paris StreetView ablation, mask ratio (0.1, 0.2] | SSIM: 0.903 -> 0.889 | Hurt |
| use LBAM instead of PConv | image inpainting · Places, mask ratio (0.4, 0.5] | PSNR (dB): 21.38 -> 21.66 | Improved |
Closest papers by meaning
- Image Inpainting for Irregular Holes Using Partial Convolutions (ECCV 2018 · similarity 0.90)
- Deep Learning-Based Image and Video Inpainting: A Survey (IJCV 2024 · similarity 0.89)
- Learning Pyramid-Context Encoder Network for High-Quality Image Inpainting (CVPR 2019 · similarity 0.89)
- Generative Image Inpainting With Contextual Attention (CVPR 2018 · similarity 0.89)
- Localization of Deep Inpainting Using High-Pass Fully Convolutional Network (ICCV 2019 · similarity 0.89)
- Towards Enhanced Image Inpainting: Mitigating Unwanted Object Insertion and Preserving Color Consistency (CVPR 2025 · similarity 0.87)
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.
