Facial Emotion Detector

A simple Flask application that detects facial expressions using machine learning. It captures frames from a video stream and predicts the facial expressions.

It uses a CNN architecture which is an implementation of the model as described in this research paper.

The application has been dockerized for other devs to try out.

Run Locally

In order to run this application locally follow these steps:

• Install Docker
• Start the docker engine, and
• Run the following command to pull the image

docker pull rhythmd18/facial-emotion-detector-app:latest

• Run the application using this command

docker run -p 5000:5000 rhythmd18/facial-emotion-detector-app

That's it!

Demo

Files Description

• app.py: This is the main file that runs the Flask application. It sets up the server and routes for the API, and serves the Convolutional Neural Network (CNN) model.

• utils.py: This file contains utility functions. One function takes in a frame and detect the nearest face in it. Another function predicts the emotion of the detected face by passing it through the model.

• model.py: This file defines the architecture of the CNN model using PyTorch. It is an implementation of the model defined in this paper.

• model_weights.pth: This file contains the trained parameters of the CNN model. It is used to load the weights into the model architecture defined in model.py.

• test.py: This is a standalone file used for testing the application without using the web interface. It uses OpenCV to capture frames and then uses the model to predict emotions.

• Scripts.ipynb: Notebook file which contains the model training scripts.

• App.js: This file contains scripts that interact with the API route. It sends frames to the API and receives responses in JSON format, which are then used to update the DOM.

• index.html and styles.css: Pretty self-explanatory!

Tech Stack

Client: HTML, CSS, JavaScript

Server: Flask

CNN Architecture: PyTorch

Appendix

• Dataset
• Research Paper
• PyTorch Quickstart