Paper Contents
Abstract
This paper presents a novel real-time detection system, combining pre-trained models of state-of-the-art deep learning for emotion recognition, age and gender prediction, and object detection. The design of the age and gender-prediction system utilizes pre-trained deep learning models. Emotion detection utilizes the state-of-the-art CNN model. The system employs the state-of-the-art YOLOv8 object detector to obtain and display multiple objects in live video feeds effectively and accurately. The application uses OpenCV for the in-built webcam stream, utilizing a Streamlit-based user interface so it is interactive. Some major features include options to turn grayscale onoff, object detection enableddisabled, and predictions rendered in real-time, all with optimized performance by calculating FPS and using modular components that can be easily scaled up. Applications of this work include human-computer interaction, security systems and analysis of behavior. The proposed system has multi-detection capabilities channeled into a single package; and hence, it is a tool that is all-inclusive of multi-facet real-time analysis. The system was demonstrated to be effective in live environments, and further improvements could be achieved using acceleration through the GPU and fine-tuning the model.
Copyright
Copyright © 2024 Sudhanshu Lacca. This is an open access article distributed under the Creative Commons Attribution License.
