How can I use the example Transfer Learning Using Alexnet with Vgg16?

I tried to use 'Transfer Learning Using AlexNet' with Vgg16 but it failed to start the Training iterations. How can I use this example with Vgg16?

[netTransfer, info] = trainNetwork(augimdsTrain,layers,options);

Error using trainNetwork (line 150) GPU out of memory. Try reducing 'MiniBatchSize' using the trainingOptions function.

*Error in TL_CM_V3_Test_IM_VGG16 (line 68) [netTransfer, info] = trainNetwork(augimdsTrain,layers,options);

Caused by: Error using .* Out of memory on device. To view more detail about available memory on the GPU, use 'gpuDevice()'. If the problem persists, reset the GPU by calling 'gpuDevice(1)'.*

My gpu is CUDADevice with properties:

                      Name: 'GeForce 930MX'
                     Index: 1
         ComputeCapability: '5.0'

Answer 1

[Johannes Bergstrom]

Vgg16 requires a lot of GPU memory and you don't have very much of it. The error message says "Try reducing 'MiniBatchSize' using the trainingOptions function." Did you try that? Otherwise, I would recommend using a network that uses less memory, for example, GoogLeNet or SqueezeNet.

You can use any pretriained network available in MATLAB for transfer learning in this example: https://www.mathworks.com/help/deeplearning/examples/train-deep-learning-network-to-classify-new-images.html

For a list of pretrained networks, see https://www.mathworks.com/help/deeplearning/ug/pretrained-convolutional-neural-networks.html

[Claudio Mor]

Thank you! I tried setting the minimum 'MiniBatchSize'=1 but without success. Unfortunately, I need to use Vgg16 or a custom (shallower) miniVgg for my project; I'm going to use a Nvidia Gtx1080 8Gb, can this be enough?

Answer 2

[xu lu]

I have tried many times but failed to install Vgg16 successfully. The installation always stops when the download reaches 18%.Can you help me?