How do i deploy a neural network.

I have trained a network using the deep network designer tool. Now how do i using the information on my model?

Answer 1

[drummer]

if you had configured your training options, you deploy like this:

net = trainNetwork(imds, layers, options)

If your're doing everything within the designer, did you go to the training tab, chose your training options and clicked on the Train button?

[Nitesh Ramdin]

hello Drummer,

I am building a raspberry pi car(with 4 motors) & Pi camera to use as input device.

I have taken 2000 photos while driving around a track and labeled these either LEFT, RIGHT or FORWARD.

I have trained the folders(LEFT, RIGHT or FORWARD) using the Deep Network designer. So to answer your question, yes i did train by clicking the Train button.

I am at the piont where i have exported Trained netwrk and results.

My question is how do i use these results while on a live PI camera feed to either go LEFT, RIGHT or FORWARD.

Answer 2

[Reeno Joseph]

Hi Nitesh,

Please refer the example raspi_webcam_resnet to know more about the deployment workflow.

In the above example the pre-trained network "resnet50" is used. If you have a custom n/w as a mat file, load it using "coder.loadDeepLearningNetwork('custom_network.mat');"

Hope this helps.

-Reeno