https://github.com/gustavla/fractalnet

FractalNet: A fractal-based neural network architecture
https://github.com/gustavla/fractalnet

Last synced: about 1 month ago
JSON representation

FractalNet: A fractal-based neural network architecture

• Host: GitHub
• URL: https://github.com/gustavla/fractalnet
• Owner: gustavla
• License: bsd-3-clause
• Created: 2016-06-14T20:30:02.000Z (over 8 years ago)
• Default Branch: master
• Last Pushed: 2017-10-11T10:58:44.000Z (about 7 years ago)
• Last Synced: 2024-10-31T17:39:08.494Z (about 1 month ago)
• Language: C++
• Homepage: http://people.cs.uchicago.edu/~larsson/fractalnet/
• Size: 9.77 KB
• Stars: 151
• Watchers: 21
• Forks: 45
• Open Issues: 5
• Metadata Files:
  • Readme: README.rst
  • License: LICENSE

Awesome Lists containing this project

• awesome-image-classification - unofficial-caffe : https://github.com/gustavla/fractalnet
• awesome-image-classification - unofficial-caffe : https://github.com/gustavla/fractalnet

README

        
FractalNet

==========

A fractal-based neural network architecture:

* `Project page `__

* `arXiv paper `__

Drop-path

---------

We provide a reference implementation for the elementwise-mean layer with local

drop-path. There is still no public release of local+global, but we suggest

implementing this through tying weights. 

Caffe

~~~~~

See the ``caffe`` directory for code and more information.

Fractal pattern generation

--------------------------

Wiring up a fractal network manually would take hours, so we provide simple

Python scripts that will do it for you. See the ``generation`` directory for

code and more information.

Data augmentation

-----------------

We use a Python layer in Caffe to implement data augmentation. It is not yet

available here.

Cite

----

If this is useful to you, please consider citing us::

    @article{larsson2016fractalnet,

      title={FractalNet: Ultra-Deep Neural Networks without Residuals},

      author={Larsson, Gustav and Maire, Michael and Shakhnarovich, Gregory},

      journal={arXiv preprint arXiv:1605.07648},

      year={2016}

    }