The backbone of deeplabv3+ returned by deeplabv3plusLayers() using augument "xeption" only support "Xception-41" network architecture?

Inside the source code of function "deeplabv3plusLayer()" with augument "xception" , it appears to use function "xception()" return a pretrained "Xception-41" as the backbone of semantic segmentation network architecture. Can you provide the support for "Xception-65 and Xception71" as the backbone of deeplabv3+ networks?

Answer 1

[Sumukh]

Hi mingming,

I assume that the MATLAB version being used is MATLAB R2024a. I am also assuming that the number “41” in “Xception-41” refers to the depth of the pre-trained neural network” available in the Deep Learning Toolbox. This neural network already supports a depth of 71. You can refer to the following documentation to know more about the “xception” pre-trained neural network:

https://www.mathworks.com/help/releases/R2024a/deeplearning/ug/pretrained-convolutional-neural-networks.html#mw_45a8c0b2-26fa-48e9-905a-a7ed7b87bfc8

The command “deeplabv3plusLayers” is set to be removed in a future release. The command “deeplabv3plus” is recommended to be used. You can refer to the following documentation to learn more about the new command and how it replaces the “deeplabv3plusLayers” command:

https://www.mathworks.com/help/releases/R2024a/vision/ref/deeplabv3pluslayers.html#mw_8a462a47-08ad-494e-b10c-80611507154d

I hope this answers the query.