Expression Predictor

A deep learning project for predicting facial expressions from images using neural networks. This repository contains code for data preparation, augmentation, model training, and evaluation.

Project Structure

Expression-Predictor/
├── Expression-Predictor.ipynb        # Main Jupyter notebook for model development
├── generate_splits_with_datagen.py   # Script to split and augment dataset
├── img-data/                         # Directory containing training, validation, and testing images
│   ├── training/
│   ├── validation/
│   └── testing/
├── .gitignore
├── readme.md
└── .ipynb_checkpoints/

Features

• Data loading and preprocessing
• Dataset splitting (train/val/test)
• Data augmentation using Keras ImageDataGenerator
• Model training and evaluation
• Easily extensible for new classes or datasets

Setup

1. Clone the repository:

   git clone https://github.com/rbpata/Expression-Predictor
   cd Expression-Predictor

2. Install dependencies:

   pip install tensorflow numpy matplotlib scikit-learn

   (Add any other dependencies used in your notebook)

Data Preparation

• Place your images in the img-data/training/ directory, organized by class (e.g., happy/, sad/).
• Use the provided script to split and augment your dataset:

  python generate_splits_with_datagen.py

  This will create a generated/ directory with new train/val/test splits.

Model Architecture

The core model is a Convolutional Neural Network (CNN) built with TensorFlow/Keras:

model = tf.keras.models.Sequential([
    #1
    tf.keras.layers.Conv2D(16,(3,3),activation = 'relu',input_shape = (200,200,3)),
    tf.keras.layers.MaxPool2D(2,2),
    #2
    tf.keras.layers.Conv2D(16,(3,3),activation = 'relu'),
    tf.keras.layers.MaxPool2D(2,2),
    #3
    tf.keras.layers.Conv2D(16,(3,3),activation = 'relu'),
    tf.keras.layers.MaxPool2D(2,2),
    #4
    tf.keras.layers.Flatten(),
    #
    tf.keras.layers.Dense(1,activation="sigmoid")
])

Usage

• Open Expression-Predictor.ipynb in Jupyter Notebook or JupyterLab.
• Follow the notebook cells to:
  • Load and preprocess data
  • Build and train your model
  • Evaluate performance

Customization

• Modify generate_splits_with_datagen.py to change split ratios or augmentation settings.
• Add new classes by creating new folders in img-data/training/.