How to modify the fully connected layer for YOLO purposes?

Greetings

I have two classes and I want to use the Squeezenet network for YOLOv3 purposes. However, the last three layers in the network ('pool10', 'prob', and 'ClassificationLayer_predictions') have an ouput of 1x1x1000. How can I change the network for my purpose?

Answer 1

[Hornett]

Hi Fahmi,

You can use MATLAB's removeLayers function in order to remove a layer from a neural network, then you can add your own custom layer with the given output size of 1x1x1000 using addLayers function.

• removeLayers: https://www.mathworks.com/help/deeplearning/ref/nnet.cnn.layergraph.removelayers.html
• addLayers: https://www.mathworks.com/help/deeplearning/ref/nnet.cnn.layergraph.addlayers.html

Hope it helps!