This is a reimplementation of the paper On Data-Driven Saak Transform (https://arxiv.org/abs/1710.04176), maintained by Jiali Duan and Yueru Chen.

psnr metric: 104.294772826

psnr metric: 105.637763477

psnr metric: 105.513759179

psnr metric: 106.509034118

• [Dataset] ( Hand-written digits classification)

  • [MNIST] ( train set: 60000, 28x28. We used the same with downloaded from http://yann.lecun.com/exdb/mnist/)

• [Installation] (sklearn and Pytorch)

  • [Sklearn Installation] Refer to http://scikit-learn.org/stable/install.html)
  • [Pytorch Installation] (Refer to http://pytorch.org)
  • [Optional: Jupyter Notebook] (Refer to http://jupyter.org/install.html)

• [How to] (Forward and Inverse Transform)

  • Forward Transform: multi_stage_saak_trans
  • Inverse Transform: toy_recon(outputs,filters)

• [To-do list]

  • One-stage Saak Transform
  • Multi-stage Saak Transform
  • Inverse Transform

• [Other Code]

  • [notebook] multi-stage_saak.ipynb
  • [dataset I/O] datasets.py, utils.py

• Contact Me

Jiali Duan (Email: [email protected])

Yueru Chen (Email: [email protected])