SK-AM62A-LP: apps_python query

Hello Sajan,

For an image classification model, the output of the model is the confidence score on a per-class basis.

For instance, a model trained on imagenet dataset can recognize 1000 classes. Often there is an additional 'null' class to cover the negative case. Thus, you can an output of 1x1001. See the model here [0] and open in Netron to see the output dimensions.

You query mentions apps_python in the title, but it is not clear if you are meaning this question for a specific TI tool otherwise. I will assume you mean edgeai-gst-apps, but please clarify otherwise if needed. 

In apps_python, see the inference_pipe.py file where the model is run [1] and result passed along to post-processing [2]. For classificaiton model, this should be a numpy array with inner dimension of 1000 or 1001, if you are using a classification model in our model zoo. 

[0] https://github.com/TexasInstruments/edgeai-tensorlab/blob/main/edgeai-modelzoo/models/vision/classification/imagenet1k/edgeai-tv2/mobilenet_v2_lite_wt-v2_20231101_model.onnx.link 

[1] https://github.com/TexasInstruments/edgeai-gst-apps/blob/799dcbda54f829eb7b234bf93a5669addcc1b919/apps_python/infer_pipe.py#L110 

[2] https://github.com/TexasInstruments/edgeai-gst-apps/blob/799dcbda54f829eb7b234bf93a5669addcc1b919/apps_python/infer_pipe.py#L121

BR,
Reese