deepdream error at cuda

Question

michael scheinfeild 2019-3-24

0
链接

此问题的直接链接

https://ww2.mathworks.cn/matlabcentral/answers/452019-deepdream-error-at-cuda

回答： Maria Duarte Rosa 2019-4-5

采纳的回答： Maria Duarte Rosa

hi i want to see the deepdream image of googlenetwork pool1-norm1 it works for next layers i receive error

anyone knows what this error means ?

Error using nnet.internal.cnngpu.convolveForward2D

Unexpected error calling cuDNN: CUDNN_STATUS_BAD_PARAM.

 net_ser = SeriesNetwork(netGoogleTrained.Layers);
  layern='pool1-norm1';%5 ok
  layern='inception_3a-3x3'%16  fail
 ch=[1 2];
 I = deepDreamImage(net_ser,layern,ch);
 
 figure
for i = 1:2
    subplot(2,1,i)
    imshow(I(:,:,:,i))
end

0 个评论
显示 -2更早的评论隐藏 -2更早的评论

请先登录，再进行评论。

请先登录，再回答此问题。

Answer 1

Maria Duarte Rosa 2019-4-5

0
链接

此回答的直接链接

https://ww2.mathworks.cn/matlabcentral/answers/452019-deepdream-error-at-cuda#answer_369191

Hi Michael, thanks for your question.

deepDreamImage does not currently support DAG networks, such as GoogLeNet, and there is no way to convert a DAGNetwork into a SeriesNetwork because the latter can only be used for networks with all the layers arranged one after the other (such as AlexNet).

For other visualization techniques I suggest:

https://www.mathworks.com/help/deeplearning/examples/visualize-activations-of-a-convolutional-neural-network.html

https://www.mathworks.com/help/deeplearning/examples/investigate-network-predictions-using-class-activation-mapping.html

https://blogs.mathworks.com/deep-learning/2017/12/15/network-visualization-based-on-occlusion-sensitivity/

https://blogs.mathworks.com/deep-learning/2019/01/18/neural-network-feature-visualization/

Having said this, deepDreamImage is expected to work with DAG networks in future releases.