Skip to main content

A library to build and train neural networks in with TensorFlow for Computer Vision

Project description

[![Code Issues](](

A simple toolkit to easily apply image classification networks.

## Usage

Train a model:

$ python

Evaluate a model:

$ python

* --config=Examples/myconfig
* --name=myname

## Requirements

The following python moduls need to be installed:

* tensorflow 0.71
* numpy
* scipy
* PIL / pillow

You can do that by

$ pip install -r requirements.txt

## Development

First of all, you should install the additional development requirements:

$ pip install -r requirements-dev.txt

For development, you can avoid reinstalling the TensorVision package by adding
TensorVision to your PYTHONPATH. In my case, that is:

$ export PYTHONPATH="${PYTHONPATH}:/home/moose/GitHub/TensorVision/"

You can run the tests by

$ python test

## Workflow

### One-time stuff

* Create a for general configuration

### Each time

Each time you get a new task

#### Create JSON file

Create a json file (e.g. `cifar10_cnn.json`). It has at least the following

"model": {
"input_file": "examples/inputs/",
"architecture_file" : "examples/networks/",
"optimizer_file" : "examples/optimizer/"

#### Adjust input file

The `input_file` contains the path to a Python file. This Python file has to
have a function `inputs(hypes, q, phase, data_dir)`.

#### Adjust architecture file

The `architecture_file` contains the architecture of the network. It has to
have the following functions:

* `loss(H, logits, labels)`
* `inference(H, images, train=True)`
* `evaluation(H, logits, labels)`

#### Adjust the solver file

The `optimizer_file` contains the path to a Python file. This Python file has
to have a function `training(H, loss, global_step)`. It defines how one tries
to find a minimum of the loss function.

## Changelog

Project details

Release history Release notifications

This version
History Node


Supported by

Elastic Elastic Search Pingdom Pingdom Monitoring Google Google BigQuery Sentry Sentry Error logging CloudAMQP CloudAMQP RabbitMQ AWS AWS Cloud computing Fastly Fastly CDN DigiCert DigiCert EV certificate StatusPage StatusPage Status page