UNet | Medical Image Segmentation | Computer Vision library

Two things:

1. Please check which version of TensorFlow (TF) you are using. I believe that from 2.5, you don't need to rescale the input image to the range [0-1]. The network expects tensors from [0-255]. https://keras.io/api/applications/efficientnet/
2. Your model architecture and callbacks look all right (I am not an expert on this optimizer + loss though). Thus, I am assuming that the problem might come from your data input. Are you using ImageDataGenerator as input and for splitting the data from training and validation? If not, it might be worth a try. You can specify your validation subset and the generator will split the data for you. More info here: https://www.tensorflow.org/api_docs/python/tf/keras/preprocessing/image/ImageDataGenerator

Use train_batches in model.fit and not train_images. Also, you do not need to use repeat(), which causes an infinite dataset if you do not specify how many times you want to repeat your dataset. Regarding your labels error, try rewriting your model like this:

For the basic layers (e.g., nn.Conv, nn.Linear, etc.) the parameters are initialized by the __init__ method of the layer.
For example, look at the source code of class _ConvNd(Module) (the class from which all other convolution layers are derived). At the bottom of its __init__ it calls self.reset_parameters() which initialize the weights.

Therefore, if your nn.Module does not have any "independent" nn.Parameters, only trainable parameters inside sub-nn.Modules, when you construct your network, all weights of the sub modules are being initialized as the sub modules are constructed. That is, once you call h_model = H_model() the weights of h_model are already initialized to their default values. Calling h_model.load_state_dict(...) overrides these values to the desired pre-trained weights.

AT command shells (and Groovy shells) can be extended with shell extensions. Shell extensions implement a org.arl.fjage.shell.ShellExtension tag interface. Static methods (and attributes) of the shell extension class are made available in the shell as commands (and variables/constants). In the AT command shell, they are called using AT commands as briefly described in section 12.3 of the Unet handbook.

For example, the handbook shows an example of loading the PhysicalShellExt via:

There was a code logic mistake in the forward of Encoder

I did:

About first question test_img_norm=test_img[:,:,0][:,:,None], test_img[:,:,0] will copy first channel of image and test_img[:,:,0][:,:,None] will add one channel to it. for example if you have an image with shape (256, 256, 3), test_img_norm shape will be (256, 256, 1).

About second part of question, model.predict(test_img_other_input)[0,:,:,0] > 0.2 will give you a boolean array. For every element in output of UNet, if element is less than 0.2, output would be True, otherwise would be False. And finally .astype(np.uint8) make booleans to zero or one.

You need to convert y_true to 1-hot representation in order to apply per-class dice loss. It seems like you have tf.one_hot function that does it for you.

Once you have y_true in the same shape as y_pred, you can use your code to compute the dice score for each class separately, and then combine the scores of all classes to get the final scalar loss.

Your model expects the shape (samples, 240, 240, 155, 4), so try something like this:

You are passing 1-dimensional vectors to K.dot, while the ValueError is saying that K.dot requires arrays with 2-dimensions.

You can replace it with element-wise multiplication, i.e. intersection = K.sum(targets *inputs)