/neural-network-plant-trait-classification

Fine-grained trait classification using deep convolutional neural networks

Primary LanguagePython

Semantic Fine Grained Plant Trait Classification

Update: Everything before was awful and terrible, so I needed to fix it.

What this repository contains

.
├── file_preperation - Folder containing some clever CSV parsing
├── neural_networks - Folder containing the neural network configurations
└── README.md - This file, duh.

Installing Keras.

Getting this project set up is a fairly simple task that will only take a few minutes of your time.

The project is now built on the keras theano-based deep learning library. It's built entirely on simplicity and ease of implementation.

I found the easiest way to set this up was via pip. pip is a package management system for the python programming language.

To install pip on a Ubuntu system, simply run the following command in the command line.

sudo apt-get install python-pip python-dev build-essential

Note: This project was made using Python 2.7.6.

Now that this is set up, you're going to need to install a few dependencies first.

  1. numpy and scipy: sudo pip install scipy
  2. PIL: sudo pip install pillow
  3. pyaml: sudo pip install pyyaml
  4. theano: sudo pip install theano

And finally, you can install keras using the following command.

sudo pip install keras

After that, load the software/neuralNetworkSoftware project into Pycharm, or whatever IDE/plain text editor you use and everything should be set up. If not, feel free to email me or create a new issue.

Running on a GPU

Provided that you have CUDA set up correctly, running tests on a GPU is as easy as running a line in terminal.

THEANO_FLAGS=mode=FAST_RUN,device=gpu,floatX=float32 filename.py