Closed lightlu closed 2 years ago
@lightlu Response has been updated with a simplified solution.
The pipeline definition file needs a tweak as gavdetect cannot deal with webcam input directly. As an example update object_detection/person_vehicle_bike pipeline/pipeline.json
template as follows
videoconvert
element between uridecodebin
and gvadetect
add-empty-results=true
property to gvametaconvert for debugging
"template": ["uridecodebin name=source ! videoconvert",
" ! gvadetect model={models[object_detection][person_vehicle_bike][network]} name=detection",
" ! gvametaconvert name=metaconvert add-empty-results=true ! gvametapublish name=destination",
" ! appsink name=appsink"
],
then re-start service and start pipeline as follows
$ ./vaclient/vaclient.sh run object_detection/person_vehicle_bike v4l2:///dev/video0
@lightlu does this answer your question?
@whbruce Thank you & sorry for late reply
Let me provide patch to use webcam on /dev/video0 for samples/record_frames/ which based on this git commit:
v0.6.0-beta origin/master 2534d9b4d8487864bfdccdfcdc746ccf0f5627d7
0001-Force-samples-record_frames-to-use-dev-video0-webcam.zip
@lightlu thanks for contribution. In the short term we have updated our docs. Our next release will contain a "universal" VA Serving source so that pipelines can take media from anywhere.
Web Camera Source section is noted as TBD in customizing_pipeline_requests.md. Is v4l2src source supported now? Thank you.