Object Detection with Saved Network

Hello! I have trained couple networks such as ResNet50 and GoogLeNet with my own dataset. I want to use my saved network for object detection. Is there any way to use my saved network for object detection?

[Walter Roberson]

See https://www.mathworks.com/matlabcentral/answers/1949593-how-to-install-deep-learning-toolbox-in-matlab-runtime#comment_2713273 for an outline

[Ferhat Kangal]

Thanks for the answer but what I'm trying to do is loading an image which has different classes in one picture and detecting and labeling them. As I looked for the predict() function, I do not think it's going to do what I'm trying to do.

[Walter Roberson]

If your requirement does not involve generating an executable, then you do not have any problems provided that you have the appropriate Toolbox and Support Package installed in MATLAB.

If your requirement involves generating an executable that needs to be able to train a neural network inside the executable, then you will not be able to do that using the Mathworks toolboxes: Mathworks restricts the training facilities so that they cannot be deployed to an executable.

If your requirement involves generating an executable that needs to be able to use a previously-trained network to predict() or classify(), then the discussion I linked to describes the workflow: load() the network and use it to predict() or classify(), and getting the load() to work often involves using #%function pragmas.

[Walter Roberson]

L = predict(net, Images);
class_is_present_in_image = L > 0.95;   %95% certainty
detected_classes = {};
for mask = class_is_present_in_image
    detected_classes{end+1} = classNames(mask);
end

The end result should be a cell array, one entry per image present in Images, in which the entry contains the class names of all classes present in the image to (in this case) 95% certainty.