About EmoSense

EmoSense is a real-time facial emotion recognition system. Face detection and expression classification both run on-device with WebGL-accelerated tensors.

Model

FER-style expression CNN over 48×48 grayscale inputs, 7 output classes, paired with a TinyFaceDetector for localisation.

Performance

Inference runs asynchronously with a single job in flight, so the interface stays responsive at 25–30 FPS on typical hardware.

Privacy

Webcam frames are processed locally and are never transmitted or stored anywhere.

Emotion classes

😊Happy
😢Sad
😠Angry
😨Fear
😮Surprise
😐Neutral
🤢Disgust
😴Tired

Emotion recognition from facial expressions is probabilistic and can be affected by lighting, occlusion, head pose and demographic bias in training data. Treat results as an estimate, not a diagnosis.