Research map / Image generation / Autoregressive and latent generation
Multi-View Data Generation Without View Supervision
Mickael Chen, Ludovic Denoyer, Thierry Artières
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 |
|---|---|---|---|
| Use pair-based adversarial training with a shared content vector and separate sampled view vectors | GMV · CelebA | identity preservation AUC, generated positive vs generated negative pairs: 72.0 % -> 76.3 % | Improved |
| Use shared content with separately sampled views rather than generating a fixed multi-image bundle | GMV vs GANx8 · CelebA | identity preservation AUC, real positive vs generated negative pairs: 13.3% -> 78.7 % | Improved |
| Condition the generator on an encoded input and generate views through C-GMV | C-GMV vs CGAN · CelebA | blurry-image rate: 99.8 % -> 37.2 % | Improved |
| Use C-GMV instead of CGAN for conditional view generation | C-GMV · CelebA | unique attribute-combination rate: 9% -> 43% | Improved |
| Use C-GMV instead of Mathieu (2016) (DCGAN) | C-GMV · CelebA | identity preservation AUC, input/generated positive vs generated-pair negative: 77.3 % -> 67.2% | Hurt |
| Use C-GMV instead of CGAN | C-GMV · CelebA | identity preservation AUC, generated positive vs generated negative pairs: 100 % -> 75.1% | Mixed |
| Encode content with C-GMV | C-GMV · 3D chairs | classification accuracy: 96.7% -> 93.6% | Improved |
| Generate views from encoded content with C-GMV | C-GMV · MVC cloth | classification accuracy on generated views: 45.2 % -> 27.1% | Hurt |
| Use GMV rather than GANx2 for multi-view generation | GMV · CelebA | Bhattacharyya distance to estimated attribute distribution (D2E): 0.25 -> 0.06 | Improved |
| Generate content-view pairs with GMV rather than CGAN | GMV · CelebA | blurry-image rate: 99.8 % -> 28.9% | Improved |
Closest papers by meaning
- Generative Models as a Data Source for Multiview Representation Learning (ICLR 2022 · similarity 0.91)
- Controlling generative models with continuous factors of variations (ICLR 2020 · similarity 0.89)
- Plug & Play Generative Networks: Conditional Iterative Generation of Images in Latent Space (CVPR 2017 · similarity 0.87)
- Realize Generative Yet Complete Latent Representation for Incomplete Multi-View Learning (TPAMI 2023 · similarity 0.87)
- Conditional Generative Modeling via Learning the Latent Space (ICLR 2021 · similarity 0.87)
- Compositional Scene Understanding through Inverse Generative Modeling (ICML 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.
