ValueError: cannot reshape array of size 2097152 into shape (128,128,1)

[idata]

Hi

I am trying to run a UNET architecture (built using keras using tensorflow as backend) on myriad2 using NCSDK2.05. I have ubuntu 16.04.

I converted the keras model to tensorflow graph. When I try to use .meta file with mvncProfile (-s 12 -in=Images -on=PredictionMask/Sigmoid ) I get the following error:

Blob generated

USB: Transferring Data…Time to Execute : 231.29 msUSB: Myriad Execution FinishedTime to Execute : 189.13 msUSB: Myriad Execution FinishedTraceback (most recent call last): File "/usr/local/bin/mvNCProfile", line 159, in profile_net(args.network, args.inputnode, args.outputnode, args.nshaves, args.inputsize, args.weights, args.device_no, args.explicit_concat, args.ma2480, args.scheduler, args) File "/usr/local/bin/mvNCProfile", line 145, in profile_net timings, myriad_output = run_myriad(graph_file, args) File "/usr/local/bin/ncsdk/Controllers/MiscIO.py", line 321, in run_myriad myriad_output = myriad_output.reshape(sz[1], sz[2], sz[3])ValueError: cannot reshape array of size 2097152 into shape (128,128,1)

Any help will be appreciated.

-Regards

Baishali

[idata]

does there a concat layer at your network's end?

[idata]

hi @zufeifei my last layer is the prediction layer ((Sigmoid ) Have a look at the last 3 layers: Its a keras using tensorflow as backend)

up9 = K.layers.concatenate([up, conv1], axis=concat_axis)

conv9 = K.layers.Conv2D(name="conv9a", filters=32, **params)(up9) conv9 = K.layers.Conv2D(name="conv9b", filters=32, **params)(conv9) prediction = K.layers.Conv2D(name="PredictionMask", filters=num_chan_out, kernel_size=(1, 1), data_format=data_format, activation="sigmoid")(conv9) model = K.models.Model(inputs=[inputs], outputs=[prediction])

I did go through all the forum contents concerning your prroblem. And I saw that you tried using a reshape layer after concat. Did that help?

[idata]

@baishali

a reshape layer can help you to bypass the concat layer's problem,a concay layer sometimes cause the output's shape innormal

[idata]

hi @ zulfeifei I am not sure where to add the reshape layer as i have multiple concat layers at different levels. And also I have a doubt what size of the reshape layer should I use

[idata]

the shape of output is (128, 128,128), but not (128, 128, 1). I met the same error, where the active fn is sigmoid