Skip to content
AI & ML 2024 04 / 14

Real-Time Emotion Recognition

A CNN that reads facial expressions live from a webcam — wrapped in a friendly desktop app.

Built with

Language

  • Python

Skills & tools

  • TensorFlow
  • Keras
  • OpenCV
  • PySide6
  • Deep learning
Source on GitHub (opens in a new tab)
Illustration of a wireframe human head surrounded by data overlays

Deep-learning model built with TensorFlow and Keras that classifies seven emotions in real time, with a PySide6 interface for live detection, session capture and evaluation.

Goal

Detect and classify human emotions from facial expressions — in real time, from an ordinary webcam, through an interface anyone can use.

Pipeline

  1. Dataset — scripted download and preparation of a public facial-expression dataset.
  2. Model — a convolutional neural network built with TensorFlow / Keras.
  3. Training & evaluation — dedicated scripts with reproducible metrics.
  4. Application — a PySide6 (Qt) desktop app with live detection that saves each session and the frames it classified correctly.

I later benchmarked five different approaches on the same problem — from a majority-class baseline to an augmented CNN — in Expression Model Benchmark.