Research map / Image generation / Autoregressive and latent generation
Image Shape Manipulation From a Single Augmented Training Sample
Yael Vinker, Eliahu Horwitz, Nir Zabari, Yedid Hoshen
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 |
|---|---|---|---|
| replace crop-and-flip augmentation with TPS augmentation (segmentation primitive) (ablation) | Pix2PixHD single image · Cityscapes videos (avg of 16) | LPIPS: 0.342 -> 0.216 | Improved |
| replace crop-and-flip augmentation with TPS augmentation (segmentation primitive) (ablation) | Pix2PixHD single image · Cityscapes videos (avg of 16) | SIFID: 0.292 -> 0.127 | Improved |
| add edge map to segmentation primitive (combined primitive) (ablation) | DeepSIM Seg TPS · Cityscapes videos (avg of 16) | LPIPS: 0.216 -> 0.134 | Improved |
| add edge map to segmentation primitive (combined primitive) (ablation) | DeepSIM Seg TPS · Cityscapes videos (avg of 16) | SIFID: 0.127 -> 0.104 | Improved |
| replace crop-and-flip with TPS augmentation (ablation) | Pix2PixHD single image · LRS2 sequence S1 | LPIPS: 0.44 -> 0.12 | Improved |
| replace crop-and-flip with TPS augmentation (ablation) | Pix2PixHD single image · LRS2 sequence S5 | SIFID: 0.44 -> 0.06 | Improved |
| add VGG perceptual loss (ablation) | DeepSIM without VGG loss · LRS2 sequence S1 | LPIPS: 0.14 -> 0.12 | Improved |
| add VGG perceptual loss (ablation) | DeepSIM without VGG loss · LRS2 sequence S2 | SIFID: 0.11 -> 0.12 | Mixed |
| add VGG perceptual loss (ablation) | DeepSIM without VGG loss · LRS2 sequence S4 | LPIPS: 0.28 -> 0.22 | Improved |
| add cGAN loss on top of VGG perceptual loss (ablation) | DeepSIM VGG-only · Cars image (high resolution) | perceptual quality | Improved |
| train single-image DeepSIM instead of SinGAN | SinGAN · 256x256 face image | training time (minutes): 72 -> 49 | Improved |
Earlier methods this paper compares with
Named baselines from earlier accepted papers in this paper's results tables: better than 1 of 1 on most shared results.
- SinGAN — SinGAN: Learning a Generative Model From a Single Natural Image (ICCV 2019) · better on 1 of 1 shared result
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
- Drop the GAN: In Defense of Patches Nearest Neighbors As Single Image Generative Models (CVPR 2022 · similarity 0.86)
- Learning Hierarchical Semantic Image Manipulation through Structured Representations (NeurIPS 2018 · similarity 0.86)
- Learning to Manipulate Individual Objects in an Image (CVPR 2020 · similarity 0.85)
- Learning to Manipulate Artistic Images (AAAI 2024 · similarity 0.85)
- A Generative Approach at the Instance-Level for Image Segmentation Under Limited Training Data Conditions (Student Abstract) (AAAI 2025 · similarity 0.84)
- Prompt Augmentation for Self-supervised Text-guided Image Manipulation (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.
