Skip to content
/ DFF Public
forked from edocollins/DFF

Deep Feature Factorization

Notifications You must be signed in to change notification settings

artcmd/DFF

 
 

Folders and files

NameName
Last commit message
Last commit date

Latest commit

 

History

9 Commits
 
 
 
 
 
 
 
 
 
 
 
 
 
 

Repository files navigation

http://img.shields.io/badge/arXiv-1806.10206-blue.svg?style=flat

Deep Feature Factorization (DFF)

DFF is the application of non-negative matrix faxtorization (NMF) to the ReLU feature activations of a deep neural network. In the case of CNNs trained on images, the resulting factors decompose an image batch into semenatic parts with a high degree of invariance to complex transformations.

This implementation relies on Pytorch and includes a GPU implementation of NMF with multiplicative updates (Lee and Seung, 2001).

@InProceedings{collins2018,
author = {Collins, Edo and Achanta, Radhakrishna and Susstrunk, Sabine}, title = {Deep Feature Factorization For Concept Discovery}, booktitle = {The European Conference on Computer Vision (ECCV)}, year = {2018}

}

About

Deep Feature Factorization

Resources

Stars

Watchers

Forks

Releases

No releases published

Packages

No packages published

Languages

  • Jupyter Notebook 99.7%
  • Python 0.3%