Deepstream ssd caffemodel

hello,

I am wondering if I can use initLibNvInferPlugins(&gLogger, “”) from Tensorrt5 to initialize plugin in Deepstream rather than write plugin on my own。 If the answer is yes, then how can I do it。

thanks in advance。

Firstly let’s for example uff model, someone layer which tensorRT does not support.

when convert tensorFlow PB model to uff model, we need config.py. It can change the layer name.

initLibNvInferPlugins() defines several plugins and registers them to a global plugin list.
But we don’t provide API to get the layers names. I can list here,
GridAnchor_TRT
NMS_TRT
Reorg_TRT
Region_TRT
Clip_TRT
LReLU_TRT
PriorBox_TRT
Normalize_TRT
RPROI_TRT
TensorRT uff parser gets the layer name from uff model and finds the layer definition in the global plugin list, and build the network.

For caffemodel parser, it’s the same as uff parser. Make sure layer name should be matched.