Project Goal

The primary objective of this project is to develop Machine Learning operations project for a deep learning-based artificial intelligence model that classifies various types of sports images. The model will be given a sports related image and it will output the name of the sports it’s related to. The purpose is to apply different number of coding practices to organizate, scale, monitor and deploy the machine learning model in a production setting.

Frameworks and Integration

Framework	Purpose / Usage
Git and GitHub	Code Versioning
TIMM	Pytorch based image models
DVC for Google Cloud	Data Versioning and Sharing
Conda	Environment Management
Python 3.10	Coding language
Pytorch 2.1.2	Deep Learning freamwork
VSCode and VSCode Debugger	Code Editor and Debuger
Cookiecutter	Project template
Wandb	Experiment monitoring (and hyperparameter optimization sweeping)
Ruff	Linter, make code PEP8 compliant
Docker	Create shareable environment
Pytorch-lightning	Reduce boilerplate Code
More will come…	…

Data

The initial dataset for training our model is the Sports image classification. This dataset contains 10.283 labeled images divided in two substes. The training subset contains 8227 files and the test subset contains 2056 files.

In order to have a version control of the data and make the repo lighter, dvc is going to be used. To get the data, just make sure that dvc is installed in your machine. If is not, you can do it like this:

pip install dvc

And in this case you would also need to run the following command as we are using Google Drive to store the data.

pip install "dvc[gdrive]"

Alternatively, installing the packages in requirements-dev.txt will also get dvc working, among other things. This is done with the following instruction:

pip install -r "requirements-dev.txt"

Models

The core model we expect to use is RexNet. This model is selected due to its efficiency and high accuracy in image classification tasks. We will adapt and train this model on our chosen dataset, tuning it to achieve optimal performance in sports image classification.

Conclusion

Coming soon...

Progress tracking

You can find the project progress here

Meme of the project

Installation:

Clone the repo to your local machine.
Create conda environment from requirements.txt:

Create a virtual environment:

conda create -n mlops python=3.10

Activate the environment using the following command:

conda activate mlops

Install libraries from the requirements file:

pip install -r requirements.txt

Run Model

Train the model with:

python src/main.py -c config/main.yaml fit

Project structure

The directory structure of the project looks like this:

├── Makefile             <- Makefile with convenience commands like `make data` or `make train`
├── README.md            <- The top-level README for developers using this project.
├── data
│   ├── processed        <- The final, canonical data sets for modeling.
│   └── raw              <- The original, immutable data dump.
│
├── docs                 <- Documentation folder
│   │
│   ├── index.md         <- Homepage for your documentation
│   │
│   ├── mkdocs.yml       <- Configuration file for mkdocs
│   │
│   └── source/          <- Source directory for documentation files
│
├── models               <- Trained and serialized models, model predictions, or model summaries
│
├── notebooks            <- Jupyter notebooks.
│
├── pyproject.toml       <- Project configuration file
│
├── reports              <- Generated analysis as HTML, PDF, LaTeX, etc.
│   └── figures          <- Generated graphics and figures to be used in reporting
│
├── requirements.txt     <- The requirements file for reproducing the analysis environment
|
├── requirements_dev.txt <- The requirements file for reproducing the analysis environment
│
├── tests                <- Test files
│
├── src  <- Source code for use in this project.
│   │
│   ├── __init__.py      <- Makes folder a Python module
│   │
│   ├── data             <- Scripts to download or generate data
│   │   ├── __init__.py
│   │   └── make_dataset.py
│   │
│   ├── models           <- model implementations, training script and prediction script
│   │   ├── __init__.py
│   │   ├── model.py
│   │
│   ├── visualization    <- Scripts to create exploratory and results oriented visualizations
│   │   ├── __init__.py
│   │   └── visualize.py
│   ├── train_model.py   <- script for training the model
│   └── predict_model.py <- script for predicting from a model
│
└── LICENSE              <- Open-source license if one is chosen

Created using mlops_template, a cookiecutter template for getting started with Machine Learning Operations (MLOps).

hmhauter/mlops_exam_project