Research map / Image generation / Editing and style transfer
Fine-Grained Image-to-Image Transformation Towards Visual Recognition
Wei Xiong, Yutong He, Yixuan Zhang, Wenhan Luo, Lin Ma, Jiebo Luo
Check 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 |
|---|---|---|---|
| augment few-shot training data with Ours-generated images rather than no generated data | ResNet18 · CompCars 20-way 5-shot | accuracy: 29.77 -> 52.44 | Improved |
| augment few-shot training data with Ours-generated images rather than no generated data | ResNet18 · CompCars 20-way 10-shot | accuracy: 55.86 -> 66.93 | Improved |
| use CR-GAN-generated images for augmentation instead of no augmentation | ResNet18 · CompCars 20-way 5-shot | accuracy: 29.77 -> 27.61 | Hurt |
| use CR-GAN-generated images for augmentation instead of no augmentation | ResNet18 · CompCars 20-way 10-shot | accuracy: 55.86 -> 39.67 | Hurt |
| add CNC at 28×28 to the vanilla model (ablation) | Ours generator · CompCars 20 categories | top-1 accuracy: 48.59 -> 55.05 | Improved |
| add CNC at 28×28 to the vanilla model (ablation) | Ours generator · CompCars 50 categories | top-1 accuracy: 36.20 -> 42.24 | Improved |
| add CNC at 28×28 to the vanilla model (ablation) | Ours generator · CompCars 80 categories | top-1 accuracy: 27.27 -> 34.68 | Improved |
| use constrained instead of global nonalignment matching at 28×28 (ablation) | Ours generator · CompCars 20 categories | top-1 accuracy: 53.12 -> 55.05 | Improved |
| use constrained instead of global nonalignment matching at 28×28 (ablation) | Ours generator · CompCars 50 categories | top-1 accuracy: 38.30 -> 42.24 | Improved |
| use constrained instead of global nonalignment matching at 28×28 (ablation) | Ours generator · CompCars 80 categories | top-1 accuracy: 30.40 -> 34.68 | Improved |
| add AIM rather than omit identity modulation, with CNC(28) (ablation) | Ours generator · CompCars 20 categories | top-1 accuracy: 55.05 -> 56.13 | Improved |
| add AIM rather than omit identity modulation, with CNC(28) (ablation) | Ours generator · CompCars 50 categories | top-1 accuracy: 42.24 -> 43.87 | Improved |
Closest papers by meaning
- CVAE-GAN: Fine-Grained Image Generation Through Asymmetric Training (ICCV 2017 · similarity 0.87)
- Reinforcing Generated Images via Meta-Learning for One-Shot Fine-Grained Visual Recognition (TPAMI 2022 · similarity 0.86)
- 3D Guided Fine-Grained Face Manipulation (CVPR 2019 · similarity 0.86)
- TransFG: A Transformer Architecture for Fine-Grained Recognition (AAAI 2022 · similarity 0.85)
- Progressive Adversarial Networks for Fine-Grained Domain Adaptation (CVPR 2020 · similarity 0.85)
- Coarse-to-Fine Latent Diffusion for Pose-Guided Person Image Synthesis (CVPR 2024 · similarity 0.83)
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.
