Audio Classification inference on Jetson Nsno

I have an ONNX audio classification model that came from Matlab audio toolbox. I want to run inference on Jeston Nano with TensorRT. The audio file is *.wav, a bunch of numbers. There are directions for the image classification, object detection, but not for the audio classification. Please point me to a solution.

Hi @ashmadan01, you would need to implement a program (presumably either with the TensorRT Python or C++ API) that loads your ONNX model, performs any pre-processing of the audio, calls the TensorRT engine with it, and gets the maximum-likelihood output classification. Some of this is specific to the kind of model that Matlab exported, and the input/output layers that it expects. The image classification and object detection examples had to implement something similar underneath to support models like ResNet, SSD, YOLO, ect.

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