Object detection using coral TPu

Question

ahmad 2024-3-30

0
链接

此问题的直接链接

https://ww2.mathworks.cn/matlabcentral/answers/2100761-object-detection-using-coral-tpu

回答： Shubham 2024-3-31

IMG_20240327_110241.jpg

Can matlab support coral TPu. I want to detect object using coral screenshot attached. I trained faster rccn on matlab can I used it in Coral TPu to detect object.if possible than how?

0 个评论
显示 -2更早的评论隐藏 -2更早的评论

请先登录，再进行评论。

请先登录，再回答此问题。

Answer 1

Shubham 2024-3-31

0
链接

此回答的直接链接

https://ww2.mathworks.cn/matlabcentral/answers/2100761-object-detection-using-coral-tpu#answer_1433931

Hey Ahmad,

As per my understanding, coral TPU only supports "TensorFlow Lite" models. Since you have already trained a "Faster R-CNN" model in MATLAB, export your model along with weights and architecture. Please refer to the following documentation for exporting a deep neural network to "TensorFlow":

https://www.mathworks.com/help/deeplearning/ref/exportnetworktotensorflow.html

You may also refer to the File Exchange submission by MathWorks for converting a network or a layergraph as a "TensorFlow" model:

https://www.mathworks.com/matlabcentral/fileexchange/64649-deep-learning-toolbox-converter-for-tensorflow-models

Convert the "TensorFlow" model into "TensorFlow Lite" using "TensorFlow Lite" Converter and compile it using "Edge TCU compiler".

After that you can call python in MATLAB using "pyrun" function:

https://www.mathworks.com/help/matlab/ref/pyrun.html

You can also import the "TensorFlow Lite" model in MATLAB using "loadTFLiteModel": https://www.mathworks.com/help/deeplearning/ref/loadtflitemodel.html

You may also use the Interface for the "TensorFlow Lite" by MathWorks as provided on the File Exchange:

https://www.mathworks.com/matlabcentral/fileexchange/106215-deep-learning-toolbox-interface-for-tensorflow-lite

I hope this helps!