Research map / Image generation / GANs
Efficient Generation of Structured Objects with Constrained Adversarial Networks
Luca Di Liello, Pierfrancesco Ardino, Jacopo Gobbi, Paolo Morettin, Stefano Teso, Andrea Passerini
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 semantic loss constraint term to training | WGAN · SMB pipes | validity (%): 47.6 -> 83.2 | Improved |
| Add semantic loss constraint term | WGAN · SMB pipes | average pipe tiles per level: 7.8 -> 6.4 | Hurt |
| Add semantic loss constraint term | WGAN · SMB pipes | L1 norm diversity: 0.0115 -> 0.0110 | No clear effect |
| Add semantic loss with learned reachability embedding | WGAN · SMB playable mario-1-3 | validity (%): 9.80% -> 71.60% | Improved |
| Add semantic loss with learned reachability embedding | WGAN · SMB playable mario-3-3 | validity (%): 13.00% -> 62.30% | Improved |
| Use CAN instead of GAN + CMA-ES search | WGAN · SMB playable mario-3-3 | validity (%): 64.20% -> 62.30% | No clear effect |
| Add conditional semantic loss to MolGAN | MolGAN · QM9 molecules | diversity: 91.0 -> 98.8 | Improved |
| Add conditional semantic loss to MolGAN | MolGAN · QM9 molecules | logP: 65.0 -> 73.6 | Improved |
| Add conditional semantic loss to MolGAN | MolGAN · QM9 molecules | validity: 97.4 -> 96.6 | No clear effect |
| Add conditional semantic loss to MolGAN | MolGAN · QM9 molecules | uniqueness: 2.4 -> 2.5 | No clear effect |
| Add conditional semantic loss to MolGAN | MolGAN · QM9 molecules | SA: 84.0 -> 90.7 | Improved |
| Activate semantic loss after 5000-epoch bootstrap and ramp weight linearly (ablation) | CAN · SMB pipes | pipe tile retention | Improved |
Closest papers by meaning
- Dist-GAN: An Improved GAN using Distance Constraints (ECCV 2018 · similarity 0.88)
- Constrained Generative Adversarial Networks for Interactive Image Generation (CVPR 2019 · similarity 0.87)
- Structure-preserving GANs (ICML 2022 · similarity 0.86)
- MGAN: Training Generative Adversarial Nets with Multiple Generators (ICLR 2018 · similarity 0.86)
- Structured Generative Adversarial Networks (NeurIPS 2017 · similarity 0.85)
- Escaping from Collapsing Modes in a Constrained Space (ECCV 2018 · similarity 0.85)
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.
