ValueError: Cannot feed value of shape (4, 480, 480, 3) for Tensor 'y:0', which has shape '(?, ?, ?, 2)'

[Gumballing]

Describe the bug A clear and concise description of what the bug is. The train image of mine is 4804803,the src image and mask image in same.When training there is a error: The train command is: from tf_unet import unet,util,image_util

preparing data loading

data_provider=image_util.ImageDataProvider("./train/*.jpg")

setup & training

net=unet.Unet(layers=3,features_root=64,channels=3,n_class=2) trainer=unet.Trainer(net) path=trainer.train(data_provider,"./train-out",training_iters=32,epochs=100) To Reproduce Steps to reproduce the behavior:

Expected behavior A clear and concise description of what you expected to happen.

Screenshots If applicable, add screenshots to help explain your problem.

Additional context Add any other context about the problem here.

[Gumballing]

What could be the cause of this problem? what should I do? Has anyone else encountered this problem?

[Gumballing]

Find the problem： The value of the n_class required by the network after loading the image is 3, which is 2 in my previous command. Now I used the train command net=unet.Unet(layers=3,features_root=64,channels=3,n_class=3).Training is normal. But in my mask image I only label one type of target, the other is the background, why does the n_class need to be set to 3?

[obelling]

I have the same problem,can you tell me how to solve this?

[Gumballing]

I have the same problem,can you tell me how to solve this?

You need to check the annotated images in your training data. The mask is best in grayscale.

[obelling]

@Gumballing While the annotated image is in grayscale(one channel),n_class is 2 ,I fear that training will has error like: Cannot feed value of shape (4, 256, 256, 1) for Tensor 'y:0', which has shape '(?, ?, ?, 2)'

[obelling]

I had try convert the labels to gray and successfully train the model.Thank you very mach!

[Gumballing]

I had try convert the labels to gray and successfully train the model.Thank you very mach!

no thank. I have some problem in testing the model.If you tested success,I will need to consult your test code.