sketch priors, CVPR 2003
motivation
- 不依赖于prior training
- 在待测试的图像上训练,利用图像内部的冗余信息。
- 在acquisition process未知的图像上超过已有的SOTA supervised 的CNN和unsupervised方法。
- “supervised” for any method that trains on externally supplied examples (even if their generation does not require manual
labelling).
- “supervised” for any method that trains on externally supplied examples (even if their generation does not require manual
contribution
- 设计了