Research map / Label-efficient and robust learning / Self-supervised and contrastive
GAN-Supervised Dense Visual Alignment
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 GANgealing instead of the strongest listed baseline | SPair-71K Bicycle correspondence · SPair-71K Bicycle test | PCK-Transfer@α_bbox=0.1: 34.7 -> 37.5 | Helped |
| use GANgealing instead of the strongest listed baseline | SPair-71K Cat correspondence · SPair-71K Cat test | PCK-Transfer@α_bbox=0.1: 66.5 -> 67.0 | Helped |
| use GANgealing instead of the strongest listed baseline | SPair-71K Dog correspondence · SPair-71K Dog test | PCK-Transfer@α_bbox=0.1: 24.3 -> 23.1 | Hurt |
| use GANgealing instead of the strongest listed baseline | SPair-71K TV correspondence · SPair-71K TV test | PCK-Transfer@α_bbox=0.1: 58.0 -> 57.9 | Hurt |
| use GANgealing instead of the strongest correspondence-supervised baseline | CUB correspondence · CUB validation | PCK-Transfer@0.1: 58.5 -> 57.5 | Hurt |
| learn the target latent rather than fixing it to the average (ablation) | GANgealing · SPair-71K Cats | PCK@α_bbox=0.1: 10.6 -> 67.2 | Helped |
| constrain target optimization to 1 PCA coefficient instead of 512 (ablation) | GANgealing · SPair-71K Cats | PCK@α_bbox=0.1: 0.34 -> 67.2 | Helped |
| set the style-mixing cutoff to 4 rather than 5 (ablation) | GANgealing · SPair-71K Cats | PCK@α_bbox=0.1: 67.2 -> 60.5 | Hurt |
| set the style-mixing cutoff to 6 rather than 5 (ablation) | GANgealing · SPair-71K Cats | PCK@α_bbox=0.1: 67.2 -> 65.0 | Hurt |
| remove style mixing from per-input targets (ablation) | GANgealing · SPair-71K Cats | PCK@α_bbox=0.1: 67.2 -> 25.9 | Hurt |
| remove style mixing and use LPIPS loss (ablation) | GANgealing · SPair-71K Cats | PCK@α_bbox=0.1: 67.0 -> 7.74 | Hurt |
| remove total variation flow regularization (ablation) | GANgealing · SPair-71K Cats | PCK@α_bbox=0.1: 67.2 -> 59.0 | Hurt |
Closest papers by meaning
- CoordGAN: Self-Supervised Dense Correspondences Emerge From GANs (CVPR 2022 · similarity 0.90)
- DenseReg: Fully Convolutional Dense Shape Regression In-The-Wild (CVPR 2017 · similarity 0.87)
- PandA: Unsupervised Learning of Parts and Appearances in the Feature Maps of GANs (ICLR 2023 · similarity 0.87)
- LARGE: Latent-Based Regression Through GAN Semantics (CVPR 2022 · similarity 0.87)
- Ensembling Off-the-Shelf Models for GAN Training (CVPR 2022 · similarity 0.87)
- E$^2$GAN: Efficient Training of Efficient GANs for Image-to-Image Translation (ICML 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.