Research map / Detection and segmentation / Open-set and few-shot detection
Structure-Guided Image Completion With Image-Level and Object-Level Semantic Discriminators
Haitian Zheng, Zhe Lin, Jingwan Lu, Scott D. Cohen, Eli Shechtman, Connelly Barnes, Jianming Zhang, Qing Liu, Sohrab Amirghodsi, Yuqian Zhou, Jiebo Luo
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 |
|---|---|---|---|
| add perceptual loss to adversarial-only training (ablation) | conditional CM-GAN baseline · Places2-person (CoModGAN masks) | FID: 10.5587 -> 2.8470 | Improved |
| add perceptual loss to adversarial-only training (ablation) | conditional CM-GAN baseline · Places2-person (object masks) | FID: 9.3800 -> 2.7246 | Improved |
| add image-level semantic discriminator (ablation) | adv. + perc. model · Places2-person (CoModGAN masks) | FID: 2.8470 -> 2.2705 | Improved |
| add image-level semantic discriminator (ablation) | adv. + perc. model · Places2-person (object masks) | FID: 2.7246 -> 2.1636 | Improved |
| add object-level discriminator alone (ablation) | adv. + perc. model · Places2-person (CoModGAN masks) | FID: 2.8470 -> 2.8100 | No clear effect |
| add object-level discriminator alone (ablation) | adv. + perc. model · Places2-person (object masks) | FID: 2.7246 -> 2.3477 | Improved |
| add object-level discriminator to semantic discriminator (ablation) | adv. + perc. + sem. D · Places2-person (CoModGAN masks) | FID: 2.2705 -> 2.0720 | Improved |
| add object-level discriminator to semantic discriminator (ablation) | adv. + perc. + sem. D · Places2-person (CoModGAN masks) | P-IDS (%): 22.30 -> 25.90 | Improved |
| use instance segmentation guidance instead of semantic segmentation (ablation) | full model · Places2-person (CoModGAN masks) | FID: 2.3860 -> 2.0720 | Improved |
| use Canny edge instead of instance edge guidance | full model · Places2-person (CoModGAN masks) | FID: 2.6909 -> 1.6736 | Improved |
| use our model instead of ControlNet+RePaint | ControlNet*+RePaint · COCO-Stuff segmentation-guided (object masks) | FID: 9.9252 -> 10.1776 | Hurt |
| use our one-step GAN instead of diffusion ControlNet+RePaint | ControlNet+RePaint · inference time | inference time: 3 min 25 sec -> 0.279 sec | Improved |
Closest papers by meaning
- Foreground-Aware Image Inpainting (CVPR 2019 · similarity 0.88)
- Learning Hierarchical Semantic Image Manipulation through Structured Representations (NeurIPS 2018 · similarity 0.87)
- Coherent Semantic Attention for Image Inpainting (ICCV 2019 · similarity 0.87)
- ForkNet: Multi-Branch Volumetric Semantic Completion From a Single Depth Image (ICCV 2019 · similarity 0.87)
- SESAME: Semantic Editing of Scenes by Adding, Manipulating or Erasing Objects (ECCV 2020 · similarity 0.87)
- Paint by Inpaint: Learning to Add Image Objects by Removing Them First (CVPR 2025 · 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.
