Display the confidence score (from the output binding) using DeepStream

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

Hello, I have an image classification model with the following bindings:

Binding: onnx::Softmax_531, Data Type: DataType.FLOAT
Binding: output, Data Type: DataType.FLOAT

The model outputs two results:

output – This represents the final result, which is a probability.
onnx::Softmax_531 – This is the result before passing through the Softmax layer.

The softmax layer transforms the result from onnx::Softmax_531 into the probabilities available in the output binding.

However, I want to display the confidence score (probability) directly from the model, which is provided by the output binding. Currently, the code displays the result from onnx::Softmax_531, but I need the probability output from output.

Here’s my current code:
label_info = pyds.glist_get_nvds_label_info(l_label.data)
confidence = label_info.result_prob
if classifier_id == 3:
gendre3 = str(label_info.result_label)
confidence_yolo = confidence

How can I correctly display the confidence score (from the output binding) using DeepStream?

0 INPUT kFLOAT images 3x224x224
1 OUTPUT kFLOAT onnx::Softmax_531 2
2 OUTPUT kFLOAT output 2

2

and How to disable the default post-processing in DeepStream?

3

Hi @amine.ghamgui.anavid ,
Request you to raise this on Deepstream Forum.