Computer Vision (CV) is an interdisciplinary field of Artificial Intelligence (AI), which is concerned with the embedding of human visual capabilities in a computerized system. The main thrust, essentially, of CV is to generate an "intelligent" high-level description of the world for a given scene, such that when interfaced with other thought processes can elicit, ultimately, appropriate action. In this talk we will review several central CV tasks and traditional approaches taken for handling these tasks for over 50 years. Noting the limited performance of standard methods applied, we briefly survey the evolution of artificial neural networks (ANN) during this extended period, and focus, specifically, on the ongoing revolutionary performance of deep learning (DL) techniques for the above CV tasks during the past few years. In particular, we provide also an overview of our DL activities, in the context of CV, at Bar-Ilan University. Finally, we discuss future research and development challenges in CV in light of further employment of prospective DL innovations.
展开▼
