Research map / Label-efficient and robust learning / Semi-supervised learning
UGC: Unified GAN Compression for Efficient Image-to-Image Translation
Yuxi Ren, Jie Wu, Peng Zhang, Manlin Zhang, Xuefeng Xiao, Qian He, Rui Wang, Min Zheng, Xin Pan
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 unlabeled data (semi-supervised learning) in NAS supernet training (ablation) | UGC UNet · Edges2Shoes 10% labels | FID: 38.16 -> 34.80 | Improved |
| add unlabeled data in online distillation fine-tuning (ablation) | UGC UNet · Edges2Shoes 25% labels | FID: 30.1 -> 27.90 | Improved |
| add unlabeled data in both NAS and fine-tuning (ablation) | UGC UNet · Cityscapes 25% labels | mIoU: 45.77 -> 48.5 | Improved |
| use NAS-searched architecture instead of fixed architecture (ablation) | UGC UNet 1.2G MACs · Cityscapes 25% labels | mIoU: 44.64 -> 48.50 | Improved |
| use NAS-searched architecture instead of fixed architecture (ablation) | UGC UNet 0.8G MACs · Edges2Shoes 10% labels | FID: 66.30 -> 37.80 | Improved |
| use UGC compression vs OMGD at similar MACs | Pix2Pix ResNet · Edges2Shoes 25% labels | FID: 32.30 -> 21.43 | Improved |
| compress with UGC at 25% labels vs original with 100% labels | Pix2Pix ResNet · Edges2Shoes | FID: 24.18 -> 21.43 | Improved |
| UGC compression of GauGAN at 25% labels | GauGAN · Cityscapes 25% labels | mIoU: 55.54 -> 61.53 | Improved |
| UGC at 50% labels vs OMGD at 100% labels | Pix2Pix UNet · Edges2Shoes | FID: 25.00 -> 23.25 | Improved |
Closest papers by meaning
- GAN Compression: Efficient Architectures for Interactive Conditional GANs (CVPR 2020 · similarity 0.92)
- Compressing Image-to-Image Translation GANs Using Local Density Structures on Their Learned Manifold (AAAI 2024 · similarity 0.91)
- Content-Aware GAN Compression (CVPR 2021 · similarity 0.90)
- Co-Evolutionary Compression for Unpaired Image Translation (ICCV 2019 · similarity 0.90)
- E$^2$GAN: Efficient Training of Efficient GANs for Image-to-Image Translation (ICML 2024 · similarity 0.89)
- Rethinking the Paradigm of Content Constraints in Unpaired Image-to-Image Translation (AAAI 2024 · similarity 0.88)
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.
