Research map / Adversarial robustness and security / Attacks on text and graphs
Adversarial Self-Defense for Cycle-Consistent GANs
Read the paperCheck 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 |
|---|---|---|---|
| add noise to the translation before reconstruction (noisy cycle loss) | CycleGAN · GTA V | IoU p2p: 0.203 -> 0.230 | Helped |
| add noise to the translation before reconstruction | CycleGAN · GTA V | reconstruction honesty (RH): 27.434 -> 9.166 | Helped |
| add noise to the translation before reconstruction | CycleGAN · GTA V | sensitivity to noise (SN): 446.924 -> 94.150 | Helped |
| add guess discriminator loss | CycleGAN · GTA V | IoU segm: 0.157 -> 0.169 | Helped |
| add guess discriminator loss | CycleGAN · GTA V | reconstruction honesty (RH): 27.434 -> 11.380 | Helped |
| add guess discriminator loss | CycleGAN · GTA V | sensitivity to noise (SN): 446.924 -> 212.589 | Helped |
| add noise to the translation before reconstruction | CycleGAN · Google Maps | IoU segm: 0.175 -> 0.187 | Helped |
| add noise to the translation before reconstruction | CycleGAN · Google Maps | sensitivity to noise (SN): 251.192 -> 222.176 | Helped |
| add guess discriminator loss | CycleGAN · Google Maps | reconstruction honesty (RH): 21.775 -> 7.467 | Helped |
| add noise defense | CycleGAN · SynAction | MSE: 32.547 -> 22.182 | Helped |
| add guess loss defense | CycleGAN · SynAction | MSE: 32.547 -> 23.565 | Helped |
| use smaller cycle-loss weights proportional to domain relation (lambda_A=5, lambda_B=3) without defenses (ablation) | CycleGAN · GTA V | translation reliability | Hurt |
Closest papers by meaning
- On Translation and Reconstruction Guarantees of the Cycle-Consistent Generative Adversarial Networks (NeurIPS 2022 · similarity 0.90)
- Distilling Portable Generative Adversarial Networks for Image Translation (AAAI 2020 · similarity 0.89)
- Theoretical Analysis of Image-to-Image Translation with Adversarial Learning (ICML 2018 · similarity 0.89)
- Breaking the Cycle - Colleagues Are All You Need (CVPR 2020 · similarity 0.88)
- Connecting the Digital and Physical World: Improving the Robustness of Adversarial Attacks (AAAI 2019 · similarity 0.88)
- StegoGAN: Leveraging Steganography for Non-Bijective Image-to-Image Translation (CVPR 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.