Paper
AET vs. AED: Unsupervised Representation Learning by Auto-Encoding Transformations rather than Data
The success of deep neural networks often relies on a large amount of labeled examples, which can be difficult to obtain in many real scenarios. To address this challenge, unsupervised methods are strongly preferred for training neural networks without usi
arxiv.org
Code
github.com/maple-research-lab/AET
maple-research-lab/AET
Auto-Encoding Transformations (AETv1), CVPR 2019. Contribute to maple-research-lab/AET development by creating an account on GitHub.
github.com
AET: The Proposed Approach
특정한 transformation을 이용해 image를 변환하고 변환 전 image x 와 변환 후 image t(x)를 encoding한다.
이를 이용해 decoding하여 transformation을 예측하게 하고, 실제 transformation t와 예측된 transformation t(hat)
을 비교하여 loss-function을 구성한다.
