Description

Introduction

Face expression detection can be used for various applications such as baby monitoring system with marketing analysis. In this project, Nvidia’s low-cost GPU board Jetson Nano is used as the main core. It uses “Haar cascade frontal face” and “emotion” model to detect the face and to recognize the emotions.

Abstract

In this project, the USB camera is interfaced with Jetson Nano, “Haar cascade frontal face” model is used to detect the face first. So that it can apply the “emotion” model to the specific face area of the frame to detect the emotions. In this experiment emotions like Normal, Sad, Anger and happy can be detected.

Existing system

In the existing system, the same application is developed in Matlab platform which is purely on image processing applied and deployed in Laptop/Computer.

Proposed system

In this proposed system, the same application is deployed in hardware which can be implanted in any places where it requires to detect the emotions. Here Jetson Nano is used, which has high video processing frame rate. Most suitable for real-time implementation.

Demo video

Connection description

USB camera is interfaced with Nano, as alternative code can be easily modified to use Mobile camera or any other CSI Interfaceable camera for better resolution.

Project description

Nvidia Nano is booted with Ubuntu Operating system with every library like OpenCV, Keras and Numpy is installed. It uses Python programming. Initially for face detection “Haarcascade frontal face ” model is loaded to detect the face from the complete frame. Then to detect emotions “emotions” model is loaded and applies that model to the detected face part. Face emotion is detected in real-time and labeled to represent the emotion. This can be implemented in Baby monitoring to intimate that the baby is abnormal.

Hardware Required

• Nvidia Jetson Nano
• 32GB SD card
• USB Camera / Mobile camera

Software required

• OS for Nvidia nano
• SD card formatter
• Etcher

Result

Real-time face emotion recognition is done in Nvidia Jetson Nano, which is most suitable for the real-time video-based application