Hello, I have deployed my own neural network on the Raspberry Pi 5 AI kit and it can run normally, but the measured speed is very slow (the image processing frame rate is very low).
Based on the parameter count of my neural network and some examples provided by Hailo officially, I feel that the resources of the Hailo 8 chip are not fully utilized. Therefore.
I would like to ask if there is any way to check the utilization rate of the Hailo 8 chip or any methods to improve the operation speed.
I recommend to have a look at the following post for some generic points about this topic.
Hailo Community - My model runs slower than expected
Additionally, I recommend to have a look at the profiler in the Hailo AI Software Suite. The Profiler report will give you a detailed look into the used resources.
Once you converted your model successfully, you can use the Performance Mode of the Dataflow Compiler. The Performance Mode can be used to compile the model to the highest possible resource utilization, to maximize performance (FPS). Expect the compilation time to increase dramatically.
Hi @zhao_yj
You can use hailortcli monitor command to see the utilization rate. See A Comprehensive Guide to Building a Face Recognition System - #35 by Conor_Roche for an example usage and what you can expect to see.