Hello Experts,
CC: @Honey_Patouceul @DaneLLL @amycao @kayccc @icornejo.a @AastaLLL @dusty_nv @neuezeal @Jeffli
Unlike other developments the sample applications provided for the Deepstream SDK is full of loading and linking plugins.
1. Is that the way it is expected to be developed
2. Is there any Deepstream App dev guide
3. How to add custom models into Deepstream SDK
hi techguyz:
1:yes, Deepstream based on gstreamer which is streamed pipeline that means you could extend it’s functions with plugin. for example ,utilize the GPU(dGPU,iGPU) or other HW processor(VIC ,PVA and so on ) of NV product to speed up you pipeline
2:these guide can be referenced
https://docs.nvidia.com/metropolis/#page/DeepStream_Plugin_Manual/deepstream_plugin_faq.html#wwpID0EGHA
3: custom models can be add with TRT transferred (from caffe, uff,onnx)to TRT engine with trtexec tools, or on deepstream you can let deepstream transferred for your custom models directly(caffe, uff, onnx) ,more details code you can find from example code DIR in deepsteam install path