@ThisIsLamer commented on GitHub (Jan 5, 2024):
I made all the changes and this is the result I got
It seems that two cores are loaded again, but not fully, and they manage a framerate of…
@NickM-27 commented on GitHub (Jan 5, 2024):
worth mentioning that the 2688x1520 stream is being scaled down to 1280x720 here which uses software on the rockchip so that is likely part of the usage
@MarcA711 commented on GitHub (Jan 5, 2024):
Hi,
all devices seem to be present. So, hardware acceleration should work now.
Could you try out this docker-compose.yml:
@ThisIsLamer commented on GitHub (Jan 5, 2024):
I would also like to add that after launching the stable version of the application, which has no support for Rockchip chips, the CPU load is…
@ThisIsLamer commented on GitHub (Jan 4, 2024):
I understand that this question might be slightly off-topic for Frigate, but I am encountering difficulties in enabling hardware acceleration on…
@ThisIsLamer commented on GitHub (Jan 4, 2024):
You can also see from the log that there were problems launching npu
the container is running in privileged mode
docker-compose.yml
@NickM-27 commented on GitHub (Jan 4, 2024):
CC @MarcA711
@MarcA711 commented on GitHub (Jan 5, 2024):
Hey ThisIsLamer,
What does ls /dev/dri return? Does it include renderD128 renderD129? (renderD129 is your VPU used to decode frames from your…
@ThisIsLamer commented on GitHub (Jan 5, 2024):
Hello, MarcA711.
I tried running the image you recommended, attempted to write it to emmc using RKDevTools, and to a flash drive using…
@Dinth commented on GitHub (Jan 4, 2024):
I have updated it to match the reolink config in the doc:
@ThisIsLamer commented on GitHub (Jan 4, 2024):
I also found this error, if you run…
@daernsinstantfortress commented on GitHub (Jan 4, 2024):
Thanks - I'll look at this later today and will report back.
@NickM-27 commented on GitHub (Jan 4, 2024):
Please read the docs, you are not using the GPU for decoding and not using the npu for object detection so it's no surprise CPU usage is high
@belmont commented on GitHub (Jan 7, 2024):
0.13 has extremely good detection ! Night detection never worked well enough with frigate for me until 0.13. Kudos to devs! Side note i use nvidia GPU…
@NickM-27 commented on GitHub (Jan 4, 2024):
localhost:8443 is incorrect, should be 8554
@NickM-27 commented on GitHub (Jan 4, 2024):
See https://deploy-preview-6262--frigate-docs.netlify.app/configuration/hardware_acceleration#rockchip-platform for GPU
https://deploy-preview-6262--…
@daernsinstantfortress commented on GitHub (Jan 9, 2024):
I've continued to monitor and the CPU load increase has steadied out to not be as high as initially feared, especially given the…
@Dinth commented on GitHub (Jan 4, 2024):
Thanks! I dont know how this happened or where the 8443 came from :(
@blakeblackshear commented on GitHub (Jan 4, 2024):
In almost all cases, this has been due to the increased sensitivity of motion detection in 0.13.
Default values have been changed for…
@ThisIsLamer commented on GitHub (Jan 4, 2024):
I noticed that for my camera, ffmpeg is launched with the following parameters: