An AI-Powered Computer Vision Module for Social Interactive Agents
F. Xavier Gaya-Morey, Cristina Manresa-Yee, José María Buades Rubio · 2024
Social interactive agents play a crucial role in various domains, providing intelligent assistance in healthcare, entertainment, and education settings. Recent advancements in Artificial Intelligence (AI) have shown promising potential to enhance the autonomy of these agents. However, the lack of standardization in their development often results in the creation of complex functionalities that are challenging to transfer across different platforms. In this study, we introduce a general-purpose AI-powered computer vision module designed to address this challenge. Our module features a modular structure that enables easy scalability and integration into diverse environments. Currently supporting seven tasks, including face and person detection, facial recognition, facial expression recognition, facial landmarks estimation, age and gender estimation, and background subtraction, the module offers up to 21 computer vision methods. Additionally, we integrate explainability functionalities to enhance user trust in the system. Moving forward, we aim to expand the module by adding new tasks and methods to meet evolving needs. Our goal is to streamline the integration of AI capabilities into social interactive agents, simplifying their development and enhancing their utility across various applications.