The teams from the Center for Hardware Signal Processing, established through collaboration between the Palace of Science and the Faculty of Electrical Engineering at the University of Belgrade, achieved exceptional results at this year's IEEE Low Power Computer Vision Challenge – one of the leading international events dedicated to addressing challenges in computer vision using artificial intelligence, held from March 1 to 31, 2025.
This year's competition featured three challenges, with the shared goal of developing advanced AI algorithms capable of running on hardware-constrained devices, such as the Snapdragon 8 Gen 3 mobile chips used in modern smartphones. In such conditions, key factors for success were efficiency, optimization, and real-time performance.
Two teams were formed within the Center for Hardware Signal Processing, participating in the first and third challenges and achieving outstanding results.
The ETF - Amigo team, consisting of Marko Staletović, Maša Đurić, Tara Stepanović, Viktor Živković, Teodora Trmčić, Filip Došljak, Uroš Pantelić, and Nikola Drakulić, competed in the first challenge – developing a system for image recognition under various lighting conditions and formats, including artificially generated images. Despite this being their first participation in such a competition, the team achieved an excellent result, securing third place, behind renowned research centers from China and South Korea, as well as industry representatives.
In the third challenge, the ETF - Perceptron team, consisting of Matija Hadži Stanković, Miroslav Čolić, and Todor Ostojić, participated. Their task was particularly technically demanding – scene depth estimation from a single image (known as Monocular Depth Estimation), where the algorithm had to function in real-time, with a minimum of 30 frames per second. Despite the challenges, the team performed excellently, finishing in fifth place in the final ranking.
The supervisor of both teams was Assistant Professor Dragomir El Mezeni from the Department of Electronics and Digital Systems at the Faculty of Electrical Engineering, who heads the Center for Hardware Signal Processing at the Palace of Science, which is part of the SPARC group of the Faculty of Electrical Engineering.
These successful performances represent a significant contribution to the promotion of domestic students and experts in the field of artificial intelligence, reaffirming the importance of supporting young talents and the development of local research capacities.
