Facial Smile Detection

Table of contents

• Overview
• Datasets
• Preprocessing
  • image preprocessing
  • image augmentation
• Model
  • Building the model
  • Training
• Results

Overview

Datasets

Preprocessing

• image preprocessing

• image augmentation

Model

Building the model

The CNN that is composed of:

• Conv2D layer with 32 filters, a kernel size of (3, 3), the relu activation function, a padding equal to same and the correct input_shape
• MaxPooling2D layer with a pool size of (2, 2)
• Conv2D layer with 64 filters, a kernel size of (3, 3), the relu activation function, and a padding equal to same
• MaxPooling2D layer with a pool size of (2, 2)
• Conv2D layer with 128 filters, a kernel size of (3, 3), the relu activation function, and a padding equal to same
• MaxPooling2D layer with a pool size of (3, 3)
• Flatten layer
• dense function with 50 neurons with the relu activation function
• dropout layer (with a rate of 0.5), to regularize the network
• dense function related to the task: binary classification > sigmoid

Training

Results