Replace model in Deepstream SDK on Jetson (Solved)

Accelerated Computing Intelligent Video Analytics DeepStream SDK
1

I have Jetson-TX2, jetpack 3.1, cuda 8.0
Deepstream works good.
Now I want to YOLO darknet model for object detection.
I read the documentation given.
Following steps are given-
The configuration parameters must be specified, as applicable, in:
[primary-gie]/[secondary-gie*]
The configuration parameters that must be specified include:
• model-file
• proto-file
• model-cache, if already generated
• labelfile-path
• mean-file, if required
• offsets, if required
• parse-func, must be set to 0, used for detectors
• parse-bbox-func-name, for detectors
• parse-bbox-lib-name, for detectors

Can I know if anybody have tried to replace the model?
Or Can I know how to do it in depth?
Thank you

2

Pharate,
For Yolo, our TRT can’t support all layer directly.
So it need IPlugin in TRT, which is not supported in our gst-nvinfer plugin.

So for this network, pls break down the whole pipeline now.
for example:
RTSP-> decode → VIC → APPSINK to get frame data, then use TRT + IPlugin to do inference.

Thanks
wayne zhu