How to get input fed to nvinfer with Python bindings

Hello, I am using DeepStream 6.2 with its official Docker image, a Tesla T4, and the Python bindings.
I noticed that my model sometimes is not performing well when using DeepStream. The model simply seems to output random data sometimes.
I want to understand better what is causing this, and I would like to be able to save the preprocessed input that is fed to the nvinfer plugin. How can I do that using Python? It looks like there are no Python bindings for NvDsPreProcessBatchMeta and I don’t know C++. Are there other way?

There is no update from you for a period, assuming this is not an issue anymore. Hence we are closing this topic. If need further support, please open a new one. Thanks

The simpler way is to directly modify our c/c++ source code: NvDsInferStatus NvDsInferContextImpl::queueInputBatch(NvDsInferContextBatchInput &batchInput) in sources\libs\nvdsinfer\nvdsinfer_context_impl.cpp.
We don’t have a python demo for preprocess link sgie now. If you need to use it now, you may need to add binding by yourself. You can refer to the link below:
We will consider adding such python demos in the future.

This topic was automatically closed 14 days after the last reply. New replies are no longer allowed.