Machine learning · Computer vision / 2025
Facial Emotion Recognition
A machine-learning project that recognizes emotions from facial-landmark geometry, comparing interpretable models with a neural-network benchmark.
Finding patterns.
Reading expression.
- Year
- 2025
- Focus
- Machine learning · Computer vision
- Tools & topics
- Python / scikit-learn / OpenCV / face_recognition
The thinking behind it.
Built with a three-person team, this project evaluates whether lightweight, landmark-based features can classify seven facial emotions without relying on raw-image classification or deep CNNs. The dataset contains more than 35,000 labeled 48×48 grayscale images.
I implemented model code as well as facial-landmark extraction, feature engineering, and data cleanup. The pipeline detects faces, derives distances, angles, and ratios from 68 landmarks, standardizes the resulting features, and evaluates Logistic Regression, Decision Tree, and MLP models with cross-validation and classification metrics.