Image recognition

Table of contents

• Installation
• Instructions
• Project motivation
• File descriptions
• Results
• Creator
• Thanks

Installation

In order to be able to execute your own python statements it should be noted that scripts are only tested on anaconda distribution 4.5.11 in combination with python 3.6.6. The scripts require additional python libraries.

Run the following commands in anaconda prompt to be able to run the scripts that are provided in this git repository.

• conda install scikit-learn
• conda install pandas
• conda install numpy
• conda install pytorch
• conda install tochvision
• conda install argparse
• conda install tqdm

Two quick start options are available:

• Download the latest release.
• Clone the repo: git clone https://github.com/FrankTub/DeepLearning.git

Instructions:

Note that you will need to download the training data yourself and place this somewhere on your machine where the python script can access it.

Train your model

Example of use:
   - python train.py data_directory              ==> where data_directory is directory where the train data is located.
   - python train.py data_dir -s save_directory  ==> -s allows you to dictate the directory where the checkpoint will be saved.
   - python train.py data_dir -a "vgg13"         ==> -a dictates the architecture of the pretrained model.
   - python train.py data_dir -g                 ==> -g enables GPU training.
   - python train.py data_dir -L 0.001           ==> -L sets the learning rate.
   - python train.py data_dir -H 512             ==> -H sets the number of hidden_layers in the classifier.
   - python train.py data_dir -E 4               ==> -E sets the number of iteration over the training set.

Make predictions

Example of use:
   - python train.py input checkpoint            ==> where input is filepath of image and checkpoint is filepath where the model is saved.
   - python train.py input checkpoint -t 4       ==> where 4 represents the number of flowers returned by this script.
   - python train.py input checkpoint -c bla.json==> where bla.json represents the file that translates the number to a flower name.
   - python train.py input checkpoint -g         ==> -g enables predictions on GPU.

Project motivation

For the first term of the nanodegree become a data scientist of Udacity I got involved in this project. I was particular interested in trying out deep learning python libraries and see what results could be achieved.

File descriptions

Within the download you'll find the following directories and files.

DeepLearning/
├── README.md
├── Image Classifier Project.ipynb # Notebook to try out deep learning python libraries
├── predict.py # Make prediction for an image
├── utility.py # Helper functions
├── train.py # Python script to train model on data directory
├── model.py # Model relation functions
└── cat_to_name.json # map index to flower name

Results

By training the classifier of a pretrained neural network I was able to make predictions on images of more than a 100 flowers with an accuracy of 85%.

Creator

Frank Tubbing

• https://github.com/FrankTub

Thanks

Thanks to Udacity for setting up the projects where we can learn cool stuff!