Visual Object Recognition - Traditional Methods Along with Deep Learning Approaches

0578865 - ÚTIA 2024 RIV LT eng C - Konferenční příspěvek (zahraniční konf.)
Flusser, Jan
Visual Object Recognition - Traditional Methods Along with Deep Learning Approaches.
14th Conference of Data Analysis Methods for Software Systems (DAMSS23). Vilnius: Vilnius University Press, 2023 - (Bernatavičienė, J.), s. 23-23. Vilnius University Proceedings, vol. 39. E-ISSN 2669-0233.
[Conference on DATA ANALYSIS METHODS for Software Systems 2023 /14./. Druskininkai (LT), 30.11.2023-02.12.2023]
Grant CEP: GA ČR GA21-03921S
Institucionální podpora: RVO:67985556
Klíčová slova: Visual Object Recognition * Deep Learning Approaches * continuous analysis of the visual field
Obor OECD: Computer hardware and architecture
http://library.utia.cas.cz/separaty/2023/ZOI/flusser-0578865.docx

The talk falls into the area of visual Artificial Intelligence (AI), particularly to image recognition by deep networks. In AI applications such as surveillance systems, autonomous robots, unmanned vehicles, drones, etc., cameras and other visual sensors form the “eyes” of the system while image recognition algorithms substitute the visual cortex of the brain. The key requirement is a continuous (possibly real-time) analysis of the visual field and, in that way, preparing the basis for decision and next action planning. The visual analysis may comprise scene segmentation, detection of objects and persons of interest, recognition of their identity and their behaviour, and even prediction of their next actions.
Trvalý link: https://hdl.handle.net/11104/0347795