TensorRT7.1 How to add ProposalPlugin in network?

AI & Data Science Deep Learning (Training & Inference) TensorRT
,
1

Hi,
Currently, I would like to construct a CNN network by tensorrt7.1 C++ API.
I want to call the “ProposalPlugin” function to help me finish my work as soon as possible, but the issue block my step.
I follow this link https://github.com/NVIDIA/TensorRT/tree/release/7.1/plugin/proposalPlugin
but I can not found the callable function in NvInferPlugin.h, and I can not use it as below method.

nvinfer1::IPluginV2* RPNROIPlugin = createProposalPlugin();
ITensor* inputTensors_RPNROI = {rpn2->getOutput(0), rpn3_sigmod->getOutput(0)};
nvinfer1::IPluginV2Layer* RPNROI = network->addPluginV2(inputTensors_RPNROI,2,*RPNROIPlugin);
assert(RPNROI != nullptr);
std::string layerName = “custom”;
RPNROI->setName(layerName.c_str());

Thanks a lot for you reply.

2

Hi @472941960,
Request you to check the below sample .

Thanks!

7

Hi,
I have review this code. https://forums.developer.nvidia.com/t/tensorrt7-1-how-to-add-proposalplugin-in-network/146600/2
But, some key point for me such as “ProposalPlugin” has been wraped and I can not know more details.

Currently, I got a trained model in mxnet and I have no chance to retrain it in TF.
So, I want to know how to call this plugin that have been implemented by NVIDIA https://github.com/NVIDIA/TensorRT/tree/release/7.1/plugin as easy as sample code in deepstream in “opt/nvidia/deepstream/deepstream-5.0/sources/objectDetector_Yolo/nvdsinfer_custom_impl_Yolo/yolo.cpp”,

nvinfer1::IPluginV2* regionPlugin = createRegionPlugin(RegionParameters);
assert(regionPlugin != nullptr);
nvinfer1::IPluginV2Layer* region = network.addPluginV2(&previous, 1, *regionPlugin);
assert(region != nullptr);

eeee, I can not call ProposalPlugin func, do not founded any definition in head file.
if this plugin is easy to call, construct a model in C++ API will be effictive and facility.

8

Hi @472941960,
Are you still facing the issue?