Flatten and Dense layers in TensorFlow

Introduction

Flatten and Dense layers help a neural network understand and learn from data by organizing it into simple shapes and making decisions.

When you want to convert a multi-dimensional image into a single list of numbers for processing.

When building a simple neural network to classify handwritten digits.

When connecting convolutional layers to fully connected layers in an image recognition model.

When you want the model to learn patterns and make predictions from features.

Syntax

TensorFlow

from tensorflow.keras.layers import Flatten, Dense

# Flatten layer example
flatten_layer = Flatten(input_shape=(height, width, channels))

# Dense layer example
dense_layer = Dense(units=number_of_neurons, activation='activation_function')

The Flatten layer changes multi-dimensional data into a flat list.

The Dense layer is a fully connected layer where each neuron connects to all inputs.

Examples

Flattens a 28x28 grayscale image into a 784-length vector.

TensorFlow

Flatten(input_shape=(28, 28, 1))

A Dense layer with 128 neurons using ReLU activation to add non-linearity.

TensorFlow

Dense(units=128, activation='relu')

A Dense layer with 10 neurons for classification into 10 classes using softmax.

TensorFlow

Dense(units=10, activation='softmax')

Sample Model

This code builds a simple neural network with Flatten and Dense layers to classify 28x28 grayscale images into 10 classes. It trains on random data for 2 epochs and shows prediction results.

TensorFlow

import tensorflow as tf
from tensorflow.keras import Sequential
from tensorflow.keras.layers import Flatten, Dense

# Create a simple model
model = Sequential([
    Flatten(input_shape=(28, 28, 1)),  # Flatten 2D input to 1D
    Dense(128, activation='relu'),  # Hidden layer
    Dense(10, activation='softmax') # Output layer for 10 classes
])

# Compile the model
model.compile(optimizer='adam', loss='sparse_categorical_crossentropy', metrics=['accuracy'])

# Create dummy data: 100 samples of 28x28 grayscale images and labels
import numpy as np
x_train = np.random.random((100, 28, 28, 1))
y_train = np.random.randint(0, 10, 100)

# Train the model for 2 epochs
history = model.fit(x_train, y_train, epochs=2, verbose=2)

# Make predictions on first 5 samples
predictions = model.predict(x_train[:5])
print('Predictions shape:', predictions.shape)
print('Predictions for first sample:', predictions[0])

Important Notes

The Flatten layer does not learn any parameters; it just reshapes data.

Dense layers learn weights and biases to find patterns in data.

Activation functions like ReLU and softmax help the model learn complex patterns and output probabilities.

Summary

Flatten turns multi-dimensional data into a flat list for Dense layers.

Dense layers connect all inputs to neurons to learn from data.

Together, they help build simple neural networks for classification tasks.