Skip to content
/ keynet Public

Key-Nets: Optical Transformation Convolutional Networks for Privacy Preserving Vision Sensors

License

Notifications You must be signed in to change notification settings

visym/keynet

Folders and files

NameName
Last commit message
Last commit date

Latest commit

 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 

Repository files navigation

Keynet

Key-Nets: Optical Transformation Convolutional Networks for Privacy Preserving Vision Sensors

Keynet

[project] [paper] [video]

Installation

python 3

# Virtualenv installation
pip3 install scipy numpy torch torchvision vipy scikit-learn xxhash numba tqdm
pip3 install -e .  
# Virtualenv installation (Linux, multi-threaded scipy/numpy)
pip3 install intel-scipy intel-numpy torch torchvision vipy scikit-learn xxhash numba tqdm
pip3 install -e .  

Quickstart

from keynet.mnist import LeNet_AvgPool
from keynet.system import PermutationKeynet

net = LeNet_AvgPool()
(sensor, knet) = PermutationKeynet(inshape=(1,28,28), net=net, do_output_encryption=False)
y = knet.forward(sensor.load('./demo/owl.jpg').encrypt().tensor())

Citation

Key-Nets: Optical Transformation Convolutional Networks for Privacy Preserving Vision Sensors
Jeffrey Byrne (Visym Labs), Brian Decann (STR), Scott Bloom (STR)
British Machine Vision Conference (BMVC) 2020, Best Paper - Runner up

@InProceedings{Byrne2020bmvc,
author = {J. Byrne and B. Decann and S. Bloom},
title = {Key-Nets: Optical Transformation Convolutional Networks for Privacy Preserving Vision Sensors},
booktitle = {British Machine Vision Conference (BMVC)},
year = {2020}
}

Acknowledgement

This material is based upon work supported by the Defense Advanced Research Projects Agency (DARPA) under Contract No. HR001119C0067.

About

Key-Nets: Optical Transformation Convolutional Networks for Privacy Preserving Vision Sensors

Resources

License

Stars

Watchers

Forks

Languages