custom object detection model based on yolov8n. Input layer 96x96.
RPI5 with AI Kit
Hailo8L
CPU Cortex-A76
This time I’m trying to stream video at the highest possible fps. When I first launched the hailo-rpi5-examples repository using my “pi camera module 3”, I saw that the maximum fps was ~30fps.
So, i decided to replace 30 with 120 in the hailo_apps_infra.gstreamer_app code. But I got only 55 fps instead of the expected 120.
I would like to know where i can switch the mode of picamera ? As i understood from specification it can works in 3 modes. So i want to use mode 1536x864p120 to get 120 fps.
Greetings, in my previous topic, I asked myself this question. when I ran the script for detection from hailo-rpi5-examples repository. But I didn’t use the rpi camera as input. I used an mp4 video instead.
So here is the results
Unfortunately, I currently don’t have the opportunity to test CPU Usage when using the rpi camera, but I’m pretty sure that it is not the main problem at the moment. I mean, I know that such a problem exists in my processor. But it looks like my fps has some invisible barrier that it can’t break through. I think my CPU has the ability to send frames at ~100 fps, but the camera itself currently acts as a bottleneck. Before my changes, it limited fps to 30, and now to 50.