An advanced deep learning system for high-accuracy facial emotion recognition, real-time webcam tracking with smooth face bounding boxes, and video analysis across 7 key emotions.
Operating with ultra-fast ONNX-accelerated inference, YuNet isolates human faces with bounding box localization, landmark anchoring, and dynamic scale normalization even in challenging ambient lighting.
Leveraging dual Convolutional Block Attention Modules (CBAM), the model extracts spatial and channel attention weights to cleanly separate complex overlapping expressions such as Surprise from Fear, and Sadness from Anger.
Turn on your webcam to see real-time emotion detection, smooth face tracking, and vibrant color-coded emotion metrics.