100 Class Image Classification Project with MLOps Practices! This project is designed to demonstrate a robust and scalable approach to building and deploying a deep learning model for image classification. With 100 different classes of sports images, this project showcases the versatility and complexity of handling a large-scale image classification task.

├── 3.jpg
├── 5.jpg
├── README.md
├── app.py
├── artifacts
|   ├── data
|   |   ├── sports-classification
|   |   |   ├── test
|   |   |   ├── train
│   │   |   └── valid
│   ├── base_model
│   │   ├── base_model.keras
│   │   └── pretrained_model.keras
│   └── training
│       └── checkpoints
│           └── model_ckpt.h5
├── config
│   ├── class_index.json
│   └── config.yaml
├── dvc.yaml
├── input_image.jpg
├── notebooks
│   ├── 01_data_trail.ipynb
│   ├── 02_base_model.ipynb
│   ├── 03_model_training.ipynb
│   └── 04_model_eval.ipynb
├── params.yaml
├── reports
│   └── score.json
├── requirements.txt
├── setup.py
├── src
│   ├── LCIC
│   │   ├── __init__.py
│   │   ├── components
│   │   │   ├── __init__.py
│   │   │   ├── data_ingestion.py
│   │   │   ├── model_builder.py
│   │   │   ├── model_evaluation.py
│   │   │   └── model_training.py
│   │   ├── config
│   │   │   ├── __init__.py
│   │   │   ├── base_model_configuration.py
│   │   │   ├── configuration.py
│   │   │   ├── model_eval_configuration.py
│   │   │   └── model_training_configuration.py
│   │   ├── constants
│   │   │   ├── __init__.py
│   │   ├── entity
│   │   │   ├── __init__.py
│   │   │   ├── base_model_entity.py
│   │   │   ├── data_config_entity.py
│   │   │   ├── model_training_entity.py
│   │   │   └── model_val_entity.py
│   │   ├── main.py
│   │   ├── pipeline
│   │   │   ├── __init__.py
│   │   │   ├── stage_01_data_ingestion.py
│   │   │   ├── stage_02_base_model_building.py
│   │   │   ├── stage_03_model_training.py
│   │   │   └── stage_04_model_eval.py
│   │   ├── prediction_service.py
│   │   └── utils
│   │       ├── __init__.py
│   │       ├── common.py
│   │       └── keras_callbacks.py
├── static
│   └── prediction.css
├── template.py
└── templates
    └── index.html

• data: Contains the datasets into training, validation, and test sets.
• models: Consists of pretrain model and custome architecture defined in model_builder.py and the training script model_training.py.
• notebooks: Jupyter notebooks for data exploration and model evaluation.
• src: All the coding implementation and helper scripts for the project.
• requirements.txt: Python dependencies required for the project.
• config.yaml: Configuration other settings for the project.
• params.yaml: Configuration of the model with hyperparameters and other settings.
• main.py: Main script to run all the stages of the project.
• app.py: For run the flask app.

1. Clone this repository:

   git clone https://github.com/sadhiin/100-class-image-classifiaction.git
   cd 100-class-image-classifaction

2. Install the dependencies:

   pip install -r requirements.txt

3. Download and organize your dataset into the data directory, following the structure mentioned above.

4. Ensure to have the model related parameters in params.yaml file.

5. Configure your info (like dataset selected pretrain model) and settings in config/config.yaml.

To train the model, run the following command:

python src/LCIC/main.py

Use dvc for tracking the stages

dvc repro

This will use the configurations specified in config/config.yaml and save the trained model in the artifacts/trained_model directory.

Automatic model evaluatoin will done on test set by running the src/LCIC/main.py or dvc repro command.

Fourther if you want's to explore the code can be found at Explore model performance by running the Jupyter notebooks in the notebooks/04_model_eval.ipynb directory:

jupyter notebook notebooks/

This ensures consistent and reproducible deployments across different environments.

This project supports CI/CD pipelines to automate testing and deployment processes. Integrate with your preferred CI/CD service for seamless automation.

Feel free to contribute by opening issues or submitting pull requests. We appreciate your feedback and collaboration!

Happy coding! 🚀