Bug: up_resBlock function should not contains Conv2D

[amir-saniyan]

Hello

It seems that model implemented incorrectly.

File: model.py Line: 57 Source: https://github.com/FENGShuanglang/2D-Vnet-Keras/blob/master/model.py#L57

Decription: The right part of the V-Net network consists of a decompresses path, so it should contains de-convolutional layers, but in the current implementation, function up_resBlock uses Conv2D. It seems that model implemented incorrectly, so Conv2D should be replaced by Conv2DTranspose.

Incorrect:

conv=PReLU()(BatchNormalization()(Conv2D(16*(2**(stage-1)), 5, activation = None, padding = 'same', kernel_initializer = 'he_normal')(conv)))

Correct:

conv=PReLU()(BatchNormalization()(Conv2DTranspose(16*(2**(stage-1)), 5, activation = None, padding = 'same', kernel_initializer = 'he_normal')(conv)))

Thanks