Machine Learning for Flower Classification

Overview

Welcome to the Flower Classification project! This project leverages machine learning techniques to classify images of flowers into various species. The primary goal is to build a robust model capable of accurately identifying different types of flowers.

Key Features

• Training Script: Utilizes the train.py script to train a machine learning model on a labeled dataset of flowers.
• Prediction Script: The predict.py script allows users to make predictions on new images using a pre-trained model.
• First-Time Setup: The first_time_setup.py script assists in downloading and preparing the dataset for training.

Dataset

The project uses the Udacity Flower dataset, containing a variety of flower species. The dataset is structured to facilitate model training, validation, and testing.

Model Architecture

The project employs the VGG (Visual Geometry Group) architecture, specifically VGG-16, as the backbone for the flower classification model. The model is trained to recognize distinct features of different flower species.

Technical Requirements

To ensure optimal functionality of this project, the following technical setup is recommended:

• Python Version: Python 3.7.3. The project is developed and tested using this specific Python version for maximum compatibility.

• Libraries and Dependencies:

  • This project relies on several key Python libraries, with specific versions listed in the requirements.txt file. This includes:
    • PyTorch 1.3.1: For deep learning functionalities.
    • NumPy 1.16.2 and Pandas 0.24.2: For numerical operations, data handling, and preprocessing.
    • Pillow 5.4.1: For image processing and manipulation.
  • To install these dependencies, use the requirements.txt file:

    pip install -r requirements.txt

• Hardware Requirements: Although the project can run on a CPU, a GPU is recommended for efficient training of deep learning models, particularly for complex or large datasets.

• Operating System: The project is compatible with major operating systems such as Windows, macOS, and Linux. Note that installation steps for Python and its libraries might vary slightly across different platforms.

Installation

Follow these steps to set up the environment and run the "Machine Learning for Flower Classification" project:

1. Clone the Repository: Begin by cloning the repository to your local machine. Use the command: git clone https://github.com/ErkanHatipoglu/AIPND_final_project_part_2.git cd AIPND_final_project_part_2

2. Set Up Python Environment: It's recommended to create a new Python environment for this project. You can use Conda or any other virtual environment manager. For Conda, use: conda create --name flower-classification python=3.7.3 conda activate flower-classification

3. Install Required Packages: Install all the required packages using the requirements.txt file. Run: pip install -r requirements.txt

4. Verify Installation: After the installation, verify that all the necessary packages are correctly installed by running: pip list

5. Run the Project: Once everything is set up, you are ready to run the project scripts like train.py or predict.py.

Usage

To use the "Machine Learning for Flower Classification" project, follow these steps:

1. Getting the Data: To train the model, You can either download the dataset from this link or navigate to the project directory and run the first_time_setup.py script. This script downloads and extracts dataset to the 'flowers' directory. Example:

   python first_time_setup.py

   python first_time_setup.py --save_dir save (data set will be downloaded and extracted to the 'save' directory)

2. Training the Model: To train the model, navigate to the project directory and run the train.py script. You can specify the dataset and other parameters as needed. Example:

   python train.py (data set shall be initially extracted to the 'flowers' directory)

   python train.py /path/to/flower_data (data set shall be initially extracted to the 'data_dir' directory)

   python train.py /path/to/flower_data --save_dir save_directory (set directory to save checkpoints)

   python train.py /path/to/flower_data --arch "vgg13" (choose architecture from vgg13, vgg16 and vgg19)

   python train.py /path/to/flower_data --learning_rate 0.01 --hidden_units [1024, 512, 256] --epochs 20 (set hyperparameters)

3. Using the Trained Model for Prediction: After training, use the predict.py script to classify new images. Provide the path to the image and the trained model checkpoint. Example: python predict.py /path/to/image.jpg checkpoint.pth

   python predict.py ( use default image 'flowers/test/1/image_06743.jpg' and root directory for checkpoint)

   python predict.py /path/to/image checkpoint (predict the image in /path/to/image using checkpoint)

   python predict.py --top_k 3 (return top K most likely classes)

   python predict.py --category_names cat_to_name.json (use a mapping of categories to real names)

   python predict.py --gpu (use GPU for inference)

4. Customizing Parameters: Both training and prediction scripts can be customized with additional command-line arguments such as setting hyperparameters for training or choosing the top K classes for prediction.

5. Reviewing Results: The training script will output the model's performance metrics, and the prediction script will display the top predicted classes along with associated probabilities.

Ensure to replace /path/to/flower_data, /path/to/image.jpg, and checkpoint.pth with the actual paths in your environment.

Acknowledgments

This project was made possible through the knowledge and skills acquired from the AI Programming with Python Nanodegree program offered by Udacity. The comprehensive curriculum and hands-on projects provided by this program were instrumental in developing the foundation required for this and future machine learning endeavors.

Special thanks to the community and resources available online, including forums, blogs, and tutorials, which have been invaluable in troubleshooting and refining the project. Among these resources, OpenAI's ChatGPT has been particularly helpful, offering guidance and support for this Readme.

This project is a testament to the collaborative nature of learning and the power of combining structured academic programs with self-directed exploration and research.