Skip to content
No description, website, or topics provided.
MATLAB Shell
Branch: master
Clone or download
Fetching latest commit…
Cannot retrieve the latest commit at this time.
Permalink
Type Name Latest commit message Commit time
Failed to load latest commit information.
hdf5creation
CRA_ft.prototxt
CRA_ft_solver.prototxt
CRA_pretrain3.prototxt
CRA_pretrain3_solver.prototxt
MissingModality_concept.jpg
README
README.md
deploy.prototxt
ft_CRA.sh
pretrain3_CRA.sh

README.md

Missing Modalities Imputation via Cascaded Residual Autoencoder (CRA)

[Project page] [CVPR'17 Paper]

Teaser

Library requirements

  • Caffe
  • (Matlab)

Data

Please create HDF5 files (see hdf4creation folder) with tags modal1, modal2, ...

Run the code

This is a part of our source code use for Multi-PIE experiment written in Caffe. The code is for a instance when the CRA is consisting of 3 RAs. Modification with different numbers of RA is straight-forward

Pretraining

pretrain3_CRA.sh

Finetunning

ft_CRA.sh

Citation

If you find this work useful, please cite our papers with the following bibtex:

@inproceedings{ tran2017missing,
  author = { Luan Tran and Xiaoming Liu and Jiayu Zhou and Rong Jin },
  title = { Missing Modalities Imputation via Cascaded Residual Autoencoder },
  booktitle = { In Proceeding of IEEE Computer Vision and Pattern Recognition },
  address = { Honolulu, HI },
  month = { July },
  year = { 2017 },
}

Contacts

If you have any questions, feel free to drop an email to tranluan@msu.edu.

You can’t perform that action at this time.