I know that deepstream can do well on video-analysis with object-detection plugins. But whether deepstream is suitable for object-detection sourced from image stream(such as opencv mat format)? If not, it that means I should recompile yolo from github source code and optimise it using tensorRT by myself instead of using deepstream SDK? (ps:I don’t know well about deepstream APIs)
We have sample "deepstream-image-decode-test " the input is jpeg file.
-
What’s your source format? Do you have to use opencv mat format?
-
You can also get raw data from opencv mat and input to streammux directly. The below is the formats streammux supports
SINK template: 'sink_%u'
Availability: On request
Capabilities:
video/x-raw(memory:NVMM)
format: { (string)NV12, (string)RGBA }
width: [ 1, 2147483647 ]
height: [ 1, 2147483647 ]
framerate: [ 0/1, 2147483647/1 ]
video/x-raw
format: { (string)NV12, (string)RGBA }
width: [ 1, 2147483647 ]
height: [ 1, 2147483647 ]
framerate: [ 0/1, 2147483647/1 ]