Flower Classification Pipeline

Project Overview

This repository contains a simple pipeline for classifying flower images. It includes utilities for downloading data from Google Drive, performing basic data cleaning, training a convolutional neural network and evaluating the result. It is meant as a starting point for experimenting with computer vision workflows and can be adapted to custom datasets by changing a few paths in the configuration file.

Requirements

• Python >=3.6
• torch==1.10.0
• torchvision==0.11.1
• tensorflow==2.7.0
• scikit-learn==0.24.2
• pandas==1.3.3
• numpy==1.21.2
• matplotlib==3.4.3
• umap-learn==0.5.2
• hdbscan==0.8.28

Setup

Clone this repository and change into its directory:

git clone <repo-url>
cd ComputerVisionStructureFlowersAI

Create a virtual environment and install the requirements:

python3 -m venv .venv
source .venv/bin/activate
pip install -r requirements.txt

You can optionally install the package in editable mode:

pip install -e .

Configuration

All dataset locations and the model save path are defined in prediction_model/config/config.py. Edit the following entries to match your environment:

• DATA_LOADING_CONFIG: Google Drive folder id and the destination directory for raw images.
• DATA_CLEANING_CONFIG: location of raw data and where the processed dataset should be written.
• TRAINING_CONFIG['model_save_path']: path where the trained model will be stored.
• EVALUATION_CONFIG['test_data_path']: directory containing test images.

Training

Run the full pipeline, which downloads data (if configured), cleans it, trains the network and finally evaluates the results:

python -m prediction_model.training_pipeline

Evaluating a Saved Model

To evaluate a model stored at the path specified in config.py using the test dataset defined in EVALUATION_CONFIG:

python -m prediction_model.evaluation.evaluate_model

Single Image Prediction

Edit prediction_model/predict.py to point image_path to your test image and then run:

python prediction_model/predict.py

Repository Structure

• prediction_model/ – package with datasets utilities, model definitions, training pipeline and configuration.
• tests/ – unit tests for the data and training modules (currently need work).
• requirements.txt – Python dependencies.
• setup.py – packaging metadata.

Limitations

The project is a work in progress. Some tests fail due to missing dependencies and several modules are incomplete. Paths in config.py must be updated before running the pipeline.

Contributions

Contributions are welcome and appreciated. If you have suggestions to improve this package or have found an error, please consider the following:

• Open an issue to discuss proposed changes or report an error.
• For direct changes, please fork the repository, make your changes, and send a pull request for review.

Before submitting a pull request, make sure your code follows the conventions established in the project and passes all existing tests. Additionally, consider adding tests for new features or fixes.

Contact

For additional questions or inquiries about this project, please don't hesitate to contact me:

• Name: Lucas Rey Braga
• Email: lucasrebra
• GitHub: lucasrebra

We appreciate your interest in this project and look forward to your contributions and feedback.