Creating APIs For Machine Learning Models with FastAPI

---

---

AI Summary

Video Summary: Turning Machine Learning Models into APIs with FastAPI

• Objective: Learn to convert machine learning models into APIs using FastAPI in Python.
• Reason: Users may not know how to use a model directly; providing an API makes it accessible.
• Tools:
  • FastAPI for creating the API.
  • Uvicorn to run the application.
  • Pillow for image processing.
  • Scikit-learn for machine learning tasks.
• Installation: Use pip to install FastAPI, Uvicorn, Pillow, and Scikit-learn.

Training the Model (train_model.py):

• Dataset: MNIST dataset for handwritten digit classification.
• Model: Random Forest Classifier.
• Process:
  • Load dataset using fetch_openml.
  • Split dataset into training and testing sets.
  • Train the model and evaluate its performance.
  • Serialize the model using pickle.

Creating the API (main.py):

• Setup: Load the serialized model and create a FastAPI app.
• Endpoint: Create a POST endpoint /predict_image to receive and process images.
• Functionality:
  • Read the uploaded file.
  • Convert to grayscale and resize if necessary.
  • Invert colors to match training data.
  • Predict using the loaded model.
  • Return the prediction as a JSON response.

Frontend (index.html):

• Interface: Simple HTML page with an input for image files and a button to classify.
• JavaScript: Handles image upload and sends a POST request to the API.
• Display: Shows the prediction result on the page.

Running the Application:

• Command: Use uvicorn to run the FastAPI app with hot reload enabled.
• Testing: Open index.html in a browser, upload an image, and view the prediction.

Conclusion:

• Outcome: Demonstrated how to expose a machine learning model as an API.
• Call to Action: Like, comment, subscribe, and hit the notification bell for future videos.

---

For more details on how to use FastAPI and other tools mentioned, you can refer to the respective documentation and tutorials within the Obsidian knowledge base.