Research map / Image generation / Image restoration
Unsupervised Transformation Learning via Convex Relaxations
Tatsunori B Hashimoto, Percy Liang, John C. Duchi
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 gradient descent refinement after sampled convex relaxation (ablation) | Sampled convex relaxation · Downsampled CelebA, single translation recovery | operator norm error | Improved |
| Use sampled convex relaxation with gradient descent instead of gradient descent with restarts | Single-transform recovery · Downsampled CelebA | operator norm error | Improved |
| Use sampled convex relaxation with gradient descent instead of EM | Single-transform recovery · Downsampled CelebA | operator norm error | Improved |
| Add SVD-based disentangling (ablation) | Multiple-transformation recovery · Downsampled CelebA | operator norm error | Improved |
| Add SVD-based disentangling (ablation) | Gradient descent baseline · Downsampled CelebA, multiple transformations | operator norm error | Improved |
| Use a five-dimensional linear transformation model learned from nearest-neighbor pairs | MNIST transformation learning · MNIST | qualitative transformation interpretability | Improved |
| Apply learned linear transformations beyond the training image distribution | MNIST-trained transformation model · Kannada handwriting | qualitative transformation transfer | Improved |
| Learn transformations with the proposed method rather than PCA | MNIST transformation visualization · MNIST | style and identity separation | Improved |
| Use the proposed method rather than InfoGAN | MNIST transformation learning · MNIST | style and identity separation | Mixed |
Closest papers by meaning
- Unsupervised Sketch to Photo Synthesis (ECCV 2020 · similarity 0.86)
- AVT: Unsupervised Learning of Transformation Equivariant Representations by Autoencoding Variational Transformations (ICCV 2019 · similarity 0.85)
- Learning Portrait Drawing with Unsupervised Parts (IJCV 2023 · similarity 0.85)
- Few-Shot Unsupervised Image-to-Image Translation (ICCV 2019 · similarity 0.84)
- Unsupervised Image-to-Image Translation with Density Changing Regularization (NeurIPS 2022 · similarity 0.84)
- Deforming Autoencoders: Unsupervised Disentangling of Shape and Appearance (ECCV 2018 · 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.
