Can't find custom caffe layer in deepstream

Accelerated Computing Intelligent Video Analytics DeepStream SDK
1

My deepStreamSDK is 4.0

Referring the example of faster rcnn, I create a custom layer as a plugin.

When running the deepstream pipeline, it says it can’t find the plugin.

nvinfer gstnvinfer.cpp:519:gst_nvinfer_logger: NvDsInferContext[UID 1]:log(): getPluginCreator could not find plugin PLUGIN_NAME version 1 namespace

All the settings follows the example of faster rcnn, and NvDsInferPluginFactoryCaffeGet is called.

How can I print all the registered plugins and debug this?

2

Hi,

Have you updated this parameter to your plugin library path:

custom-lib-path=nvdsinfer_custom_impl_fasterRCNN/libnvdsinfer_custom_impl_fasterRCNN.so

Thanks.

3

yes. I had updated the path to my own custom library.

by the way, I tried to use factoryLegacy and it works as expected.

In factoryLegacy, it implements NvDsInferPluginFactoryRuntimeGet which is called when model engine file is deserialized.

However, I can’t implement NvDsInferPluginFactoryRuntimeGet in IPluginFactoryV2, thus it can’t find the plugin when deserialization.

Do you have any idea about the deserialization part?

4

Hi,

Could you share the source code and model of your implementation so we can check it for you?
Thanks.

5

Hi OnePieceOfDeepLearning,

We haven’t heard back from you in a couple weeks, so marking this issue resolved by “accepting” this comment as answer.
Please open a new forum issue when you are ready and we’ll pick it up there.