Skip to content
A Tensorflow SqueezeNet implementation
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.
.idea
images
LICENSE
README.md
SqueezeNet.py

README.md

SqueezeNet-tf

A Tensorflow SqueezeNet implementation of the research https://arxiv.org/abs/1602.07360

What is SqueezeNet?

SqueezeNet is a convolutional neural network with 80.3% top-5 accuracy on the ImageNet dataset. This architecture aims to signifanctly reduce the size of the model.

Goals

  • Building the network
  • Training the network on imagenet
    • Currently published pre-trained weights cannot be used in TensorFlow, the goal of this project is to produce easy-to-use save files.
  • Neural Style Transfer (http://arxiv.org/abs/1508.06576) on mobile devices
    • Running on mobile devices in reasonable time without sacraficing quality too much like other feed-forward approaches.

Architecture

You can’t perform that action at this time.