Image Classification Model: Deployable Deep Learning Solution

This repository provides a complete solution for garbage classification, combining a pre-trained machine learning model with an API for easy deployment and integration.

Notebook Steps

1. Import Needed Libraries

2. Data Loading and Preprocessing

• Loads the dataset containing images of various types of garbage (e.g., plastic, glass, metal, paper).
• Performs data augmentation and preprocessing to enhance the model's robustness.

3. Model Building with ResNet50

• Utilizes the pre-trained ResNet50 model architecture, fine-tuned on the garbage classification dataset.
• Configures the model with additional layers to adapt it to the specific classification task.

4. Model Training and Evaluation

• Trains the model using the processed dataset and monitors performance through accuracy and loss metrics.
• Evaluates the model's performance on a validation set to ensure generalization and avoid overfitting.

Dataset: https://www.kaggle.com/datasets/mostafaabla/garbage-classification

API Deployment

1. FastAPI Creation

2. Dockerization

3. API Deployment with Streamlit

Note: We dockerized the API just in case we used an alternate cloud service, but you can run the API directly on Streamlit!

How to Use the API

1. Download the File

• Download file on your local device.

2. Run the FastAPI

• Write this prompt in your terminal:

  uvicorn main:app --reload
  

3. Run Streamlit

• Write this prompt in your terminal:

  streamlit run app.py

4. Upload Your Photo

• Use the provided interface to upload your photo for classification.

5. Click Predict bottom

Contact Us

nedaaelsherbini@gmail.com
eyaser@std.mans.edu.eg