DEPLOYMENT BASICS

Building a deep learning model is only the first step. Once a model is trained, the next task is to deploy it so real people can use it. Deployment means making your model available inside an app, website, mobile device, or another system. This chapter covers three essential deployment skills: saving models, using models with TensorFlow Lite, and creating a simple Flask API.

1. Saving Models

Why Save a Model?

Saves time (no need to retrain)
Makes deployment possible
Helps share models with others
Useful for backups

How Models Are Saved

TensorFlow/Keras allows you to save:

Model architecture
Weights
Optimizer settings

Code Example: Saving a Model

# Save the entire model
model.save("my_model.h5")

Loading the Model Later

from tensorflow.keras.models import load_model

model = load_model("my_model.h5")

Real Example

A model that predicts handwritten digits can be trained once and reused anytime to classify new digits without learning again.

2. Using Models with TensorFlow Lite

What Is TensorFlow Lite?

TensorFlow Lite (TFLite) allows you to run deep learning models on:

Mobile phones
Tablets
IoT devices
Microcontrollers

This Is Great For Apps Like

Face unlock
Voice recognition
Smart home devices
Offline AI apps

Why TensorFlow Lite?

Faster on mobile devices
Smaller file size
Low power usage

A. Converting a Model to TensorFlow Lite

import tensorflow as tf

# Load your trained model
model = tf.keras.models.load_model("my_model.h5")

# Convert it to TFLite
converter = tf.lite.TFLiteConverter.from_keras_model(model)
tflite_model = converter.convert()

# Save the converted model
with open("model.tflite", "wb") as f:
    f.write(tflite_model)

Explanation: Now your model is ready for mobile apps.

B. Using a TFLite Model in Mobile Apps

Android Apps

Developers load model.tflite in Java/Kotlin and run predictions on phone images or voice input.

iOS Apps

Developers use Swift and TensorFlow Lite libraries.

Example Use Cases

An AI plant disease detection app
A mobile handwritten digit recognition app
Offline translation apps

3. Flask API Basics for Deep Learning Models

What Is Flask?

Flask is a lightweight Python web framework. It allows you to create simple APIs to expose your model through the internet or a local network.

Why Use Flask?

Easy to learn
Runs locally or on servers
Lets apps send data to your model
Helps build AI-powered websites

What Is an API?

API stands for Application Programming Interface. It allows different systems to talk to each other.

Real Example

A website uploads an image → your Flask API → model predicts → sends back the result.

A. Simple Flask API Code

Step 1: Save Your Model

model.save("my_model.h5")

Step 2: Create app.py

from flask import Flask, request, jsonify
from tensorflow.keras.models import load_model
import numpy as np

app = Flask(__name__)
model = load_model("my_model.h5")

@app.route("/predict", methods=["POST"])
def predict():
    data = request.json["input"]     # Get input from user
    data = np.array(data).reshape(1, -1)
    
    prediction = model.predict(data)
    return jsonify({"prediction": prediction.tolist()})

if __name__ == "__main__":
    app.run(debug=True)

How This Works

User sends data (e.g., numbers representing features)
API sends data to the model
Model predicts the output
API returns prediction
Works on websites, mobile apps, or other systems

B. Testing the API

Use a POST Request with JSON

{
  "input": [5.1, 3.4, 1.5, 0.2]
}

The API Responds With

{
  "prediction": [[0.89]]
}

Explanation: This means the model predicts the result with 89% confidence.

4. Deployment Workflow (Step-by-Step)

Complete Deployment Process

Train the model using TensorFlow, Keras, or PyTorch
Save the model for reuse and deployment
Convert to TensorFlow Lite (optional) if running on mobile or small devices
Create a Flask API (optional) if deploying on websites or servers
Host your model on local computer, cloud (AWS, Google Cloud, Azure), or mobile app
Users send data → model predicts

5. Real-World Examples of Deployment

A. Medical AI App

Model predicts eye diseases
Doctors upload retinal images
API returns prediction instantly

B. Voice Assistant

Model runs on a phone using TensorFlow Lite
Predicts commands like "play," "pause," or "call mom"

C. E-commerce Website

Flask API recommends products
Users browse items → API suggests similar ones

DEPLOYMENT BASICS

1. Saving Models

Why Save a Model?

Saves time (no need to retrain)
Makes deployment possible
Helps share models with others
Useful for backups

How Models Are Saved

TensorFlow/Keras allows you to save:

Model architecture
Weights
Optimizer settings

Code Example: Saving a Model

# Save the entire model
model.save("my_model.h5")

Loading the Model Later

from tensorflow.keras.models import load_model

model = load_model("my_model.h5")

Real Example

A model that predicts handwritten digits can be trained once and reused anytime to classify new digits without learning again.

2. Using Models with TensorFlow Lite

What Is TensorFlow Lite?

TensorFlow Lite (TFLite) allows you to run deep learning models on:

Mobile phones
Tablets
IoT devices
Microcontrollers

This Is Great For Apps Like

Face unlock
Voice recognition
Smart home devices
Offline AI apps

Why TensorFlow Lite?

Faster on mobile devices
Smaller file size
Low power usage

A. Converting a Model to TensorFlow Lite

import tensorflow as tf

# Load your trained model
model = tf.keras.models.load_model("my_model.h5")

# Convert it to TFLite
converter = tf.lite.TFLiteConverter.from_keras_model(model)
tflite_model = converter.convert()

# Save the converted model
with open("model.tflite", "wb") as f:
    f.write(tflite_model)

Explanation: Now your model is ready for mobile apps.

B. Using a TFLite Model in Mobile Apps

Android Apps

Developers load model.tflite in Java/Kotlin and run predictions on phone images or voice input.

iOS Apps

Developers use Swift and TensorFlow Lite libraries.

Example Use Cases

An AI plant disease detection app
A mobile handwritten digit recognition app
Offline translation apps

3. Flask API Basics for Deep Learning Models

What Is Flask?

Flask is a lightweight Python web framework. It allows you to create simple APIs to expose your model through the internet or a local network.

Why Use Flask?

Easy to learn
Runs locally or on servers
Lets apps send data to your model
Helps build AI-powered websites

What Is an API?

API stands for Application Programming Interface. It allows different systems to talk to each other.

Real Example

A website uploads an image → your Flask API → model predicts → sends back the result.

A. Simple Flask API Code

Step 1: Save Your Model

model.save("my_model.h5")

Step 2: Create app.py

from flask import Flask, request, jsonify
from tensorflow.keras.models import load_model
import numpy as np

app = Flask(__name__)
model = load_model("my_model.h5")

@app.route("/predict", methods=["POST"])
def predict():
    data = request.json["input"]     # Get input from user
    data = np.array(data).reshape(1, -1)
    
    prediction = model.predict(data)
    return jsonify({"prediction": prediction.tolist()})

if __name__ == "__main__":
    app.run(debug=True)

How This Works

User sends data (e.g., numbers representing features)
API sends data to the model
Model predicts the output
API returns prediction
Works on websites, mobile apps, or other systems

B. Testing the API

Use a POST Request with JSON

{
  "input": [5.1, 3.4, 1.5, 0.2]
}

The API Responds With

{
  "prediction": [[0.89]]
}

Explanation: This means the model predicts the result with 89% confidence.

4. Deployment Workflow (Step-by-Step)

Complete Deployment Process

Train the model using TensorFlow, Keras, or PyTorch
Save the model for reuse and deployment
Convert to TensorFlow Lite (optional) if running on mobile or small devices
Create a Flask API (optional) if deploying on websites or servers
Host your model on local computer, cloud (AWS, Google Cloud, Azure), or mobile app
Users send data → model predicts

5. Real-World Examples of Deployment

A. Medical AI App

Model predicts eye diseases
Doctors upload retinal images
API returns prediction instantly

B. Voice Assistant

Model runs on a phone using TensorFlow Lite
Predicts commands like "play," "pause," or "call mom"

C. E-commerce Website

Flask API recommends products
Users browse items → API suggests similar ones

deep-learning Topics

deep-learning Tutorial

DEPLOYMENT BASICS

Why Save a Model?

How Models Are Saved

Code Example: Saving a Model

Loading the Model Later

Real Example

What Is TensorFlow Lite?

This Is Great For Apps Like

Why TensorFlow Lite?

A. Converting a Model to TensorFlow Lite

B. Using a TFLite Model in Mobile Apps

Android Apps

iOS Apps

Example Use Cases

What Is Flask?

Why Use Flask?

What Is an API?

Real Example

A. Simple Flask API Code

Step 1: Save Your Model

Step 2: Create app.py

How This Works

B. Testing the API

Use a POST Request with JSON

The API Responds With

Complete Deployment Process

A. Medical AI App

B. Voice Assistant

C. E-commerce Website

deep-learning Topics

deep-learning Tutorial

DEPLOYMENT BASICS

Why Save a Model?

How Models Are Saved

Code Example: Saving a Model

Loading the Model Later

Real Example

What Is TensorFlow Lite?

This Is Great For Apps Like

Why TensorFlow Lite?

A. Converting a Model to TensorFlow Lite

B. Using a TFLite Model in Mobile Apps

Android Apps

iOS Apps

Example Use Cases

What Is Flask?

Why Use Flask?

What Is an API?

Real Example

A. Simple Flask API Code

Step 1: Save Your Model

Step 2: Create app.py

How This Works

B. Testing the API

Use a POST Request with JSON

The API Responds With

Complete Deployment Process

A. Medical AI App

B. Voice Assistant

C. E-commerce Website