Read the buffer data recuperated from NvDsInferLayerInfo (in Python)

Accelerated Computing Intelligent Video Analytics DeepStream SDK
#1

In python I recuperated the output layer of a classifier using the same method presented here: Get class_probabilities_map from segmentation task

But I cannot read the buffer data containing the output layer values. This data is NvDsInferLayerInfo :https://docs.nvidia.com/metropolis/deepstream/sdk-api/Gst_Infer/NvDsInferLayerInfo.html

How to read this buffer data in Python?

This is the setup of my Xavier NX:

  • NVIDIA Jetson Xavier NX (Developer Kit Version)
    • Jetpack UNKNOWN [L4T 32.4.4]
    • NV Power Mode: MODE_15W_6CORE - Type: 2
    • jetson_stats.service: active
  • Libraries:
    • CUDA: 10.2.89
    • cuDNN: 8.0.0.180
    • TensorRT: 7.1.3.0
    • Visionworks: 1.6.0.501
    • OpenCV: 4.1.1 compiled CUDA: NO
    • VPI: 0.4.4
    • Vulkan: 1.2.70
Get the full vector of probabilities for every class for one classifier in DeepStream
Get the full vector of probabilities for every class for one classifier in DeepStream
#2

As asked in another topic, please provide setup info.

#3

Thanks for your reply. I added the setup information.

How to read this buffer data in Python?

#4

what’s the relation of this topic with Get the full vector of probabilities for every class for one classifier in DeepStream ?

#5

The previous topic is general. This one is specific on one method of getting the output vector.
Could you help me with this issue: How to read the buffer data from NvDsInferLayerInfo in Python?

In C, it is done like this to get the different probabilities for every class (from the example deepstream_infer_tensor_meta_test.cpp) :

    float *outputCoverageBuffer = (float *) meta->output_layers_info[0].buffer;
    float maxProbability = 0;
    bool attrFound = false;
    NvDsInferAttribute attr;

    /* Iterate through all the probabilities that the object belongs to
     * each class. Find the maximum probability and the corresponding class
     * which meets the minimum threshold. */
    for (unsigned int c = 0; c < numClasses; c++) {
      float probability = outputCoverageBuffer[c];
      if (probability > 0.51 && probability > maxProbability) {
        maxProbability = probability;
        attrFound = true;
        attr.attributeIndex = 0;
        attr.attributeValue = c;
        attr.attributeConfidence = probability;
      }
    }

In Python, the buffer is of type PyCapsule. tried to read it using the instruction:

jetson.utils.cudaToNumpy(layers_info[0].buffer)

but got this error:

Exception: jetson.utils – cudaToNumpy() failed to get input CUDA pointer from first arg (should be cudaImage or cudaMemory)

How to read this buffer?

#7

Hi @bilel_bj,
Sorry for long delay!
Do you still need help for this issue?

#8

No, thanks.

1 Like