Hi,

Recently I tried running some ROI tracking algorithms (the ones included with opencv, such as CSRT, MOSSE, etc) on an embedded ARM board and found that the tracking performance and latency is not as good as the application needs. One reason is that all these algorithms run on the CPU, which explains why they are not the fastest. Searching around, I found the VOT (Visual object tracking) challenge related to visual object tracking and it seems that most of the state-of-art trackers now use AI/ML. I guess that’s a good thing since in theory it should be easier to run these on dedicated AI accelerator hardware? Anyway, I want to try running a visual tracking algorithm on the Metis hardware to see how fast and well it can work. However, there are so many algorithms and I reckon the Metis may not support all the operators used in those networks? So before I embark on this task, i wanted to ask the Axelera experts for some guidance on what to go for and what to avoid in order to find a tracking algorithm that will be easier to run on the Metis device. I understand the question is not too specific but that’s where I am right now; some guidance to set me in the right direction will be much appreciated :)

p.s Just to clarify, I am talking about ROI tracking where you give a ROI to the algorithm and let it track whatever is inside the ROI. This is different than the object tracking commonly used with AI algorithms whereby the output of an object detector is given to a “tracker” to keep track of the same objects across the frames.