Real-Time age, gender and emotion detection using YOLOv8

The identification of age, gender, and emotion in multiple objects in an image or video stream is a complex and yet important problem for many applications such as security, health care, and human computer interaction. The current paper proposes a real-time age, gender, and emotion detection system that incorporates deep learning algorithms, in particular, the YOLOv8 model. The system employs two separate YOLO models: one for the identification of the emotion of the given video and the second one for the identification of age and gender of the subject in the video. These models are incorporated into a single pipeline where the first stage involves face detection or objects of interest and the second stage classifies the detected age, gender and emotions using pre-trained models. In real time the system is able to detect objects and classify them as well since it processes video frames taken from the webcam. The effectiveness of the proposed system is measured in terms of accuracy, running time and its ability to perform under different lighting, poses, and ethnicity. The results prove that the proposed system can accurately identify age, gender, and emotion of multiple objects and can be applied to various fields. This work shows that one may integrate emotion recognition with age-gender detection for improving the VAI (Visual Artificial Intelligence) interpretability of videos and interactions.