input dimension

[sepidhk]

my input data has the shape of (118, 224, 224, 35), 118 3D nii images, width: 224, height: 224 and depth (#of slices) 35. My question is what would be the correct order of input shape? is the order of (#images, width, height, #slices) correct? Some examples use the shape of (#images, width, height, #slices, 1). Should I set the last dimension to 1 as channel dimension?

[ZFTurbo]

Shape must be (#images, width, height, #slices, 1). Last 1 is needed.

[sepidhk]

I have followed the recommended input shape and getting error of ValueError: axes don't match array or ValueError: Input 0 of layer "model_3" is incompatible with the layer: expected shape=(None, 96, 224, 224, 35), found shape=(1, 224, 224, 35, 1).

[ZFTurbo]

You put input shape:
input_shape=(96, 224, 224, 35)

Do you have 35 channels?

If you have 1 channel use:

input_shape=(224, 224, 35, 1)

[sepidhk]

Yes, that was my thought, then I got this error: ValueError: Input 0 of layer "zero_padding3d" is incompatible with the layer: expected ndim=5, found ndim=6. Full shape received: (None, 96, 224, 224, 35, 1)

[ZFTurbo]

Shape must be: (width, height, depth, channels). What is 96?

[sepidhk]

96 is the number of images.

[ZFTurbo]

You don't need to specify it. None is used for this. It means it can be any for model

[sepidhk]

Thanks for the reply. Now with Linknet and densenet121 I have this error: ValueError: Inputs have incompatible shapes. Received shapes (56, 56, 8, 128) and (56, 56, 9, 128) and with UNet and densenet121 , I have the following error: ValueError: A Concatenate layer requires inputs with matching shapes except for the concatenation axis. Received: input_shape=[(None, 56, 56, 8, 128), (None, 56, 56, 9, 128)] - Is this because of the depth (#slices) not divisable by 32?

[ZFTurbo]

Yes. Each dimension must be divisible on 32. I think I should add warning message for this case.

[ZFTurbo]

I need to think how to make it work for any value of dimensions.

[Ek-Ram]

Hi,
I got the error : ValueError: axes don't match array. My input shape is (128,128,128, 4).
model = sm.Unet(
'resnet50',
input_shape=(128, 128, 128, 4),
encoder_weights='imagenet',
classes=3,
activation='softmax'
)
is there any solution ?

[ZFTurbo]

Can you provide exact error message?

[Ek-Ram]

Yes,
Downloading data from https://github.com/ZFTurbo/classification_models_3D/releases/download/v1.0.4/resnet50_inp_channel_3_tch_0_top_False.h5
185458688/185458584 [==============================] - 3s 0us/step
185466880/185458584 [==============================] - 3s 0us/step

ValueError Traceback (most recent call last)
in <cell line: 2>()
1 #Fit the model
----> 2 model = sm.Unet(
3 'resnet50',
4 input_shape=(128, 128, 128, 4),
5 encoder_weights='imagenet',

10 frames
/usr/local/lib/python3.10/dist-packages/segmentation_models_3D/init.py in wrapper(*args, **kwargs)
32 kwargs['models'] = _KERAS_MODELS
33 kwargs['utils'] = _KERAS_UTILS
---> 34 return func(*args, **kwargs)
35
36 return wrapper

/usr/local/lib/python3.10/dist-packages/segmentation_models_3D/models/unet.py in Unet(backbone_name, input_shape, classes, activation, weights, encoder_weights, encoder_freeze, encoder_features, decoder_block_type, decoder_filters, decoder_use_batchnorm, dropout, **kwargs)
224 'Got: {}'.format(decoder_block_type))
225
--> 226 backbone = Backbones.get_backbone(
227 backbone_name,
228 input_shape=input_shape,

/usr/local/lib/python3.10/dist-packages/segmentation_models_3D/backbones/backbones_factory.py in get_backbone(self, name, *args, **kwargs)
100 def get_backbone(self, name, *args, **kwargs):
101 model_fn, _ = self.get(name)
--> 102 model = model_fn(*args, **kwargs)
103 return model
104

/usr/local/lib/python3.10/dist-packages/classification_models_3D/models_factory.py in wrapper(*args, **kwargs)
101 modules_kwargs = self.get_kwargs()
102 new_kwargs = dict(list(kwargs.items()) + list(modules_kwargs.items()))
--> 103 return func(*args, **new_kwargs)
104
105 return wrapper

/usr/local/lib/python3.10/dist-packages/classification_models_3D/models/resnet.py in ResNet50(input_shape, input_tensor, weights, classes, stride_size, init_filters, include_top, repetitions, **kwargs)
420 **kwargs
421 ):
--> 422 return ResNet(
423 MODELS_PARAMS['resnet50'],
424 input_shape=input_shape,

/usr/local/lib/python3.10/dist-packages/classification_models_3D/models/resnet.py in ResNet(model_params, input_shape, input_tensor, include_top, pooling, classes, stride_size, init_filters, weights, repetitions, **kwargs)
338 model.load_weights(weights)
339 else:
--> 340 load_model_weights(model, model_params.model_name,
341 weights, classes, include_top, **kwargs)
342

/usr/local/lib/python3.10/dist-packages/classification_models_3D/weights.py in load_model_weights(model, model_name, dataset, classes, include_top, **kwargs)
30 )
31
---> 32 model.load_weights(weights_path)
33
34 else:

/usr/local/lib/python3.10/dist-packages/keras/utils/traceback_utils.py in error_handler(*args, **kwargs)
65 except Exception as e: # pylint: disable=broad-except
66 filtered_tb = _process_traceback_frames(e.traceback)
---> 67 raise e.with_traceback(filtered_tb) from None
68 finally:
69 del filtered_tb

/usr/local/lib/python3.10/dist-packages/numpy/core/overrides.py in transpose(*args, **kwargs)

/usr/local/lib/python3.10/dist-packages/numpy/core/fromnumeric.py in transpose(a, axes)
658
659 """
--> 660 return _wrapfunc(a, 'transpose', axes)
661
662

/usr/local/lib/python3.10/dist-packages/numpy/core/fromnumeric.py in _wrapfunc(obj, method, *args, **kwds)
55
56 try:
---> 57 return bound(*args, **kwds)
58 except TypeError:
59 # A TypeError occurs if the object does have such a method in its

ValueError: axes don't match array

[ZFTurbo]

Try encoder_weights=None

[Ek-Ram]

I tried encoder_weights=None, and it works.
but it initializes weight randomly. I prefer to use 'imagenet' weights.
is there any solution?

[ZFTurbo]

Try to do this:

pip install classification-models-3D==1.0.6

[Ek-Ram]

Try to do this:

pip install classification-models-3D==1.0.6

Unfortunately same error