[Detector Support]: OpenVino only doing CPU not GPU Intel NUC with INTEL HD Graphics #12810
-
Describe the problem you are havingI have tried every configuration under the sun to get openvino to work with Frigate on my 16gb Intel NUC with INTEL HD Graphics. Running Ubuntun 22.04 with Docker, installed the frigate docker rc.1 latest. I have tried privileged: true, and without so the below docker has been tested with and without privileged enabled. CAP_PERFMON worked to get openvino operating, but only with CPU. GPU failes everytime, only intel-vaapi is working with the GPU. I have tried AUTO as well, still no joy. I have also utilized hardware acceleration with QSV: specifically: ffmpeg: Any help or suggestions much appreciated, I am all out of ideas. Thanks! Versionfrigate:0.14.0-rc1 Frigate config filemqtt:
enabled: false
ffmpeg:
hwaccel_args: preset-vaapi
objects:
track:
- person
- car
- bird
- cat
- dog
- horse
- sheep
- cow
- bear
- backpack
- umbrella
- suitcase
- hat
- motorcycle
filters:
person:
threshold: 0.7
car:
threshold: 0.7
bird:
threshold: 0.7
cat:
threshold: 0.7
dog:
threshold: 0.7
horse:
threshold: 0.7
sheep:
threshold: 0.7
cow:
threshold: 0.7
bear:
threshold: 0.7
backpack:
threshold: 0.7
umbrella:
threshold: 0.7
suitcase:
threshold: 0.7
hat:
threshold: 0.7
motorcycle:
threshold: 0.7
detectors:
ov:
type: openvino
device: GPU
model:
path: /openvino-model/ssdlite_mobilenet_v2.xml
model:
width: 300
height: 300
input_tensor: nhwc
input_pixel_format: bgr
labelmap_path: /openvino-model/coco_91cl_bkgr.txt
detect:
stationary:
interval: 50
threshold: 50
birdseye:
enabled: true
mode: continuous
record:
enabled: true
retain:
days: 7
mode: all
events:
retain:
default: 30
mode: motion
snapshots:
enabled: true
bounding_box: true
go2rtc:
streams:
garage_cam:
ffmpeg:https://admin:(Password Removed)[email protected]:8080/stream/video/mjpeg#video=h264#ha
rdware # <- use hardware acceleration to create an h264 stream usable for other components.
cameras:
garage_cam:
ffmpeg:
inputs:
- path: rtsp:https://127.0.0.1:8554/garage_cam
roles:
- detect
- record
zones:
Garage_Driveway:
coordinates:
0.057,0.646,0.338,0.699,0.68,0.753,1,0.745,0.997,0.996,0.004,0.997,0.004,0.815,0.047,0.783,0.052,0.727
loitering_time: 0
Grass_and_Arena:
coordinates:
0,0.473,0.249,0.433,0.352,0.466,0.434,0.452,0.513,0.437,0.649,0.425,0.742,0.428,0.996,0.408,0.996,0.738,0.678,0.748,0,0.624
loitering_time: 0
review:
alerts:
required_zones:
- Garage_Driveway
- Grass_and_Arena
detections:
required_zones:
- Garage_Driveway
- Grass_and_Arena
motion:
mask:
0.004,0.009,0.003,0.341,0.049,0.406,0.134,0.364,0.159,0.329,0.188,0.305,0.228,0.326,0.273,0.339,0.301,0.335,0.297,0.401,0.303,0.437,0.355,0.459,0.444,0.442,0.5,0.421,0.515,0.399,0.492,0.367,0.501,0.262,0.512,0.177,0.562,0.134,0.621,0.137,0.687,0.163,0.692,0.244,0.73,0.295,0.773,0.32,0.833,0.291,0.898,0.26,0.999,0.298,0.998,0.003
threshold: 56
contour_area: 10
improve_contrast: 'true'
objects:
filters:
car:
mask:
0.353,0.561,0.369,0.546,0.388,0.544,0.411,0.546,0.433,0.561,0.451,0.537,0.48,0.524,0.5,0.519,0.525,0.516,0.546,0.51,0.569,0.512,0.597,0.515,0.62,0.516,0.636,0.518,0.644,0.523,0.661,0.546,0.678,0.583,0.683,0.62,0.685,0.635,0.696,0.659,0.688,0.693,0.684,0.71,0.643,0.724,0.606,0.736,0.593,0.746,0.58,0.784,0.556,0.787,0.535,0.779,0.528,0.749,0.522,0.732,0.406,0.727,0.394,0.752,0.38,0.759,0.364,0.758,0.356,0.725,0.339,0.705,0.338,0.667,0.34,0.647,0.348,0.629,0.363,0.62,0.375,0.614,0.386,0.61,0.365,0.583
version: 0.14docker-compose file or Docker CLI commandversion: "3.9"
services:
frigate:
container_name: frigate
restart: unless-stopped
image: ghcr.io/blakeblackshear/frigate:0.14.0-rc1
privileged: true
cap_add:
- CAP_PERFMON
volumes:
- ./config:/config
- ./storage:/media/frigate
- type: tmpfs # Optional: 1GB of memory, reduces SSD/SD Card wear
target: /tmp/cache
tmpfs:
size: 1000000000
devices:
- /dev/dri/renderD128:/dev/dri/renderD128
- /dev/dri/card0:/dev/dri/card0
ports:
- "5000:5000"
- "8554:8554" # RTSP feeds
environment:
LIBVA_DRIVER_NAME: i965Relevant log outputFrigate logs from Docker: s6-rc: info: service s6rc-fdholder: starting Vainfo Output
Return Code: 0
Process Output:
vainfo: VA-API version: 1.17 (libva 2.10.0)
vainfo: Driver version: Intel i965 driver for Intel(R) Broadwell - 2.4.1
vainfo: Supported profile and entrypoints
VAProfileMPEG2Simple : VAEntrypointVLD
VAProfileMPEG2Simple : VAEntrypointEncSlice
VAProfileMPEG2Main : VAEntrypointVLD
VAProfileMPEG2Main : VAEntrypointEncSlice
VAProfileH264ConstrainedBaseline: VAEntrypointVLD
VAProfileH264ConstrainedBaseline: VAEntrypointEncSlice
VAProfileH264Main : VAEntrypointVLD
VAProfileH264Main : VAEntrypointEncSlice
VAProfileH264High : VAEntrypointVLD
VAProfileH264High : VAEntrypointEncSlice
VAProfileH264MultiviewHigh : VAEntrypointVLD
VAProfileH264MultiviewHigh : VAEntrypointEncSlice
VAProfileH264StereoHigh : VAEntrypointVLD
VAProfileH264StereoHigh : VAEntrypointEncSlice
VAProfileVC1Simple : VAEntrypointVLD
VAProfileVC1Main : VAEntrypointVLD
VAProfileVC1Advanced : VAEntrypointVLD
VAProfileNone : VAEntrypointVideoProc
VAProfileJPEGBaseline : VAEntrypointVLD
VAProfileVP8Version0_3 : VAEntrypointVLDOperating systemOther Linux Install methodDocker Compose Object DetectorOpenVino Any other information that may be helpfulStandard INTEL NUC I have tried also on Dell Laptop with INTEL HD graphics and same result - there must be something I am missing, I have also set paranoid to 2 so that did not help either. |
Beta Was this translation helpful? Give feedback.
Replies: 1 comment 5 replies
-
|
you have not provided any logs showing errors, what makes you think OpenVino is not using the GPU? |
Beta Was this translation helpful? Give feedback.
-
|
Would be best to see the screenshots of the ui |
Beta Was this translation helpful? Give feedback.
-
|
Here is the main screenshot and also the system metrics screen shot:
|
Beta Was this translation helpful? Give feedback.
-
|
so what cpu do you have exactly? |
Beta Was this translation helpful? Give feedback.
-
|
CLOSING OUT : Resolved with esand work around, I do believe the frigate container will need updating to resolve long term: Also this workaround worked on my dell laptop with Intel HD and my KVM Ubuntu 22.04 with Docker with Intel HD pass through, as well on the Intel NUC which was my original submission. Just tested esands update on KVM and Dell and it worked flawless! Also for ARCH LINUX users on the host side: This worked as well, but you must enter the commands from within the docker container, not the host and for the repack as well inside the container - you will need to do that line by line and for Arch you will need to install binutils, zstd and xz-utils before running the repack commands. Then simply run deb -i *.deb in the neo folder - it will give you an error at the end, run the script again a second time and WALLA working perfect on ARCH HOST!!! cheers! Here is a link to the resolve from esand: Screenshots showing after fix on GUI and System Metrics: Smooth as silk!!! Also you will see an initial CPU spike on first 30 seconds after container launches, that goes away - so disregard!!!! You can see that spike on the 1st screen shot below:
|
Beta Was this translation helpful? Give feedback.
CLOSING OUT : Resolved with esand work around, I do believe the frigate container will need updating to resolve long term:
Also this workaround worked on my dell laptop with Intel HD and my KVM Ubuntu 22.04 with Docker with Intel HD pass through, as well on the Intel NUC which was my original submission. Just tested esands update on KVM and Dell and it worked flawless!
Also for ARCH LINUX users on the host side: This worked as well, but you must enter the commands from within the docker container, not the host and for the repack as well inside the container - you will need to do that line by line and for Arch you will need to install binutils, zstd and xz-utils before running the repack c…