diff --git a/README.md b/README.md
index f1ddbcc..f0f67fd 100644
--- a/README.md
+++ b/README.md
@@ -104,12 +104,14 @@ The compatibility test is shown as below. The checked item means this package pe
 
 ## Update reports
 
-### 0.1.5 @ 3/9/2021
+### 0.1.5 @ 3/14/2021
 
 1. Add `DecoderNet` to our standard `module` protocol.
 2. Fix some bugs of `data.h5py` and `data.preprocs`.
 3. Make `draw.setFigure` enhanced by `contextlib`.
 4. Add a title in `Readme.md`.
+5. Fix typos and bugs in `data` and `modules`.
+6. Add properties `nlayers`, `input_size` for networks in `modules`.
 
 ### 0.1.2 @ 2/27/2021
 
diff --git a/mdnc/__init__.py b/mdnc/__init__.py
index 65af74d..200df32 100644
--- a/mdnc/__init__.py
+++ b/mdnc/__init__.py
@@ -17,10 +17,13 @@
 ################################################################
 # Update reports:
 # ---------------
-# 0.1.5 @ 3/9/2021
+# 0.1.5 @ 3/14/2021
 #   1. Add DecoderNet to our standard module protocol.
 #   2. Fix some bugs of data.h5py and data.preprocs.
 #   3. Make draw.setFigure enhanced by contextlib.
+#   4. Add a title in `Readme.md`.
+#   5. Fix typos and bugs in `data` and `modules`.
+#   6. Add properties `nlayers`, `input_size` for networks in `modules`.
 # 0.1.2 @ 2/27/2021
 #   1. Fix more feature problems in `contribs.torchsummary`.
 #   2. Fix bugs and finish `data.preprocs`.
diff --git a/mdnc/data/__init__.py b/mdnc/data/__init__.py
index 289795b..90e9f49 100644
--- a/mdnc/data/__init__.py
+++ b/mdnc/data/__init__.py
@@ -29,6 +29,8 @@
 ################################################################
 # Update reports:
 # ---------------
+# 0.1.5 @ 3/14/2021
+#   1. Fix typos and bugs.
 # 0.1.2 @ 2/27/2021
 #   1. Finish preprocs.
 #      - Update the implementation of preprocs.ProcAbstract.
diff --git a/mdnc/data/__main__.py b/mdnc/data/__main__.py
index bd66b23..ae54661 100644
--- a/mdnc/data/__main__.py
+++ b/mdnc/data/__main__.py
@@ -300,7 +300,7 @@ def test_1d(self):
         plt.show()
 
     def test_2d(self):
-        x = engine.preprocs.ProcNSTScaler(parent=engine.preprocs.ProcLifter(a=1.0, parent=engine.preprocs.ProcPad(pad_width=((0, 0), (10, -10), (-10, 10)), mode='constant', constant_values=0.0)))
+        x = engine.preprocs.ProcNSTScaler(dim=2, parent=engine.preprocs.ProcLifter(a=1.0, parent=engine.preprocs.ProcPad(pad_width=((0, 0), (10, -10), (-10, 10)), mode='constant', constant_values=0.0)))
         with open(self.file_name, 'wb') as f:
             pickle.dump(x, f)
         with open(self.file_name, 'rb') as f:
diff --git a/mdnc/data/h5py.py b/mdnc/data/h5py.py
index 72de61e..224dd4c 100644
--- a/mdnc/data/h5py.py
+++ b/mdnc/data/h5py.py
@@ -783,8 +783,7 @@ def open(self, file_name, enable_read=None):
         if self.__in_context:
             raise RuntimeError('data.h5py: Should not open a file when the saver is managing a context, because there is already an opened file. Try to exit the context or create a new different saver.')
         file_name, file_ext = os.path.splitext(file_name)
-        if file_ext != '.h5':
-            file_name += '.h5'
+        file_name += '.h5'
         self.close()
         if enable_read is None:
             enable_read = self.enable_read
diff --git a/mdnc/data/preprocs.py b/mdnc/data/preprocs.py
index 63a8357..d959016 100644
--- a/mdnc/data/preprocs.py
+++ b/mdnc/data/preprocs.py
@@ -404,8 +404,8 @@ def __init__(self, shift=None, scale=None, axis=-1, inds=None, parent=None):
         self.axis = axis
 
     def preprocess(self, x):
-        xmean = np.mean(x, axis=self.axis) if self.shift is None else self.shift
-        xscale = np.amax(np.abs(x - xmean), axis=self.axis) if self.scale is None else self.scale
+        xmean = np.mean(x, axis=self.axis, keepdims=True) if self.shift is None else self.shift
+        xscale = np.amax(np.abs(x - xmean), axis=self.axis, keepdims=True) if self.scale is None else self.scale
         self.set_mem('xmean', xmean)
         self.set_mem('xscale', xscale)
         return (x - xmean) / xscale
@@ -426,7 +426,7 @@ class ProcNSTScaler(ProcAbstract):
         https://stackoverflow.com/a/49317610
     '''
 
-    def __init__(self, dim=2, kernel_length=9, epsilon=1e-6, inds=None, parent=None):
+    def __init__(self, dim, kernel_length=9, epsilon=1e-6, inds=None, parent=None):
         '''Initialization.
         Arguments:
             dim: the dimension of the input data (to be normalized).
@@ -439,6 +439,8 @@ def __init__(self, dim=2, kernel_length=9, epsilon=1e-6, inds=None, parent=None)
                     be used as the parent of the current instance.
         '''
         super().__init__(inds=inds, parent=parent)
+        if dim not in (1, 2, 3):
+            raise ValueError('data.preprocs: The argument "dim" requires to be 1, 2, or 3.')
         self.__dim = dim
         self.__kernel_length = kernel_length
         self.epsilon = epsilon
@@ -888,11 +890,11 @@ def __split_pad_width(pad_width):
                 return tuple(pad_width_), tuple(crop_width)
             else:
                 raise ValueError('data.preprocs: the crop arguments could not get separated from the pad arguments. The given arguments "pad_width" may be not valid.')
-    
+
     @property
     def pad_width(self):
-        return getattr
-    
+        return object.__getattribute__(self, '_ProcPad__pad_width_')
+
     @pad_width.setter
     def pad_width(self, value):
         self.__pad_width, self.__crop_width = self.__split_pad_width(value)
diff --git a/mdnc/data/webtools.py b/mdnc/data/webtools.py
index 98e85df..70d40c5 100644
--- a/mdnc/data/webtools.py
+++ b/mdnc/data/webtools.py
@@ -332,6 +332,18 @@ def init_set_list(file_name='web-data'):
                             'dataset_file_name_01.txt',
                             'dataset_file_name_02.txt'
                         ]
+                    },
+                    {
+                        'tag': 'test',
+                        'asset': 'test-datasets-2.tar.xz',
+                        'items': [
+                            'test_data_h5converter.h5',
+                            'test_data_h5cparser.h5',
+                            'test_data_h5cparser_seq.h5',
+                            'test_data_h5gparser.h5',
+                            'test_data_h5seqconverter1.h5',
+                            'test_data_h5seqconverter2.h5'
+                        ]
                     }
                 ],
                 'user': 'cainmagi',
diff --git a/mdnc/modules/__init__.py b/mdnc/modules/__init__.py
index 5403a85..1255ec2 100644
--- a/mdnc/modules/__init__.py
+++ b/mdnc/modules/__init__.py
@@ -16,6 +16,9 @@
 # 0.1.5 @ 3/2/2021
 #   1. Add DecoderNet to conv, resnet.
 #   2. Rename ConvNet by EncoderNet in conv, resnet.
+#   3. Fix typos and bugs.
+#   4. Add `nlayers` for all networks. Add `input_size` for
+#      decoders.
 # 0.1.0 @ 2/26/2021
 #   1. Create sub-packages: conv, resnet.
 ################################################################
diff --git a/mdnc/modules/__main__.py b/mdnc/modules/__main__.py
index 37f66b1..10cb4cd 100644
--- a/mdnc/modules/__main__.py
+++ b/mdnc/modules/__main__.py
@@ -69,6 +69,7 @@ def test_networks(self):
             print('modules.modules: Test {0}d networks.'.format(order))
             for net in self.networks:
                 test_module = net(order=order, in_planes=input_size[0])
+                print('{0} with {1} layers along its depth.'.format(type(test_module).__name__, test_module.nlayers))
                 torchsummary.summary(test_module, input_size=input_size, device='cpu')
                 del test_module
 
@@ -77,6 +78,7 @@ def test_decodernets(self):
             print('modules.modules: Test {0}d decoders.'.format(order))
             for net in self.net_decs:
                 test_module = net(order=order, in_length=2, out_size=out_size[1:])
+                print('{0} with {1} layers along its depth.'.format(type(test_module).__name__, test_module.nlayers))
                 torchsummary.summary(test_module, input_size=(2, ), device='cpu')
                 del test_module
 
diff --git a/mdnc/modules/conv.py b/mdnc/modules/conv.py
index b83fa80..5369898 100644
--- a/mdnc/modules/conv.py
+++ b/mdnc/modules/conv.py
@@ -2,18 +2,20 @@
 # -*- coding: UTF-8 -*-
 '''
 ################################################################
-# Modules - 1D convolutional network
+# Modules - convolutional network
 # @ Modern Deep Network Toolkits for pyTorch
 # Yuchen Jin @ cainmagi@gmail.com
 # Requirements: (Pay attention to version)
 #   python 3.5+
 #   pyTorch 1.0.0+
-# This module is the definition of the 1D convolutional network.
+# This module is the definition of the convolutional network.
 # The network could be initialized here and used for training
 # and processing.
 ################################################################
 '''
 
+import functools
+
 import torch
 import torch.nn as nn
 
@@ -71,7 +73,7 @@ def __init__(self, order, in_planes, out_planes, kernel_size=3, stride=1, paddin
         '''
         super().__init__()
         ConvNd = get_convnd(order=order)
-        is_stride = check_is_stride(stride)
+        is_stride = check_is_stride(stride, output_size=output_size, scaler=scaler)
         seq = []
         if normalizer == 'null':
             if (not is_stride) or scaler == 'down':
@@ -183,7 +185,7 @@ def __init__(self, order, in_planes, out_planes, hidden_planes=None,
                 _ConvModernNd(order, (in_planes + ex_planes) if i == 0 else hidden_planes, hidden_planes,
                               kernel_size=kernel_size, padding=padding, stride=1, scaler='down')
             )
-        self.conv_scale = _ConvModernNd(order, hidden_planes, out_planes,
+        self.conv_scale = _ConvModernNd(order, hidden_planes if stack_level > 1 else (in_planes + ex_planes), out_planes,
                                         kernel_size=kernel_size, padding=padding, stride=stride, scaler=scaler)
 
     @staticmethod
@@ -235,13 +237,14 @@ def __init__(self, order, channel, layers, kernel_size=3, in_planes=1, out_plane
                     of convolutional layers of a stage. The stage numer, i.e.
                     the depth of the network is the length of this list.
         Arguments (optional):
+            kernel_size: the kernel size of each block.
             in_planes: the channel number of the input data.
             out_planes: the channel number of the output data.
-            kernel_size: the kernel size of each block.
         '''
         super().__init__()
         if len(layers) < 2:
             raise ValueError('modules.conv: The argument "layers" should contain at least 2 values, but provide "{0}"'.format(layers))
+        self.__layers = layers
         ConvNd = get_convnd(order=order)
 
         ksize_e, psize_e, _ = cal_kernel_padding(kernel_size, ksize_plus=2)
@@ -273,7 +276,16 @@ def __init__(self, order, channel, layers, kernel_size=3, in_planes=1, out_plane
         self.conv_up_list.append(
             _BlockConvStkNd(order, channel, channel, hidden_planes=channel, kernel_size=ksize, padding=psize,
                             stride=1, stack_level=layers[0], ex_planes=channel, scaler='down'))
-        self.conv_final = ConvNd(channel, in_planes, kernel_size=ksize_e, stride=1, padding=psize_e, bias=True)
+        self.conv_final = ConvNd(channel, out_planes, kernel_size=ksize_e, stride=1, padding=psize_e, bias=True)
+
+    @property
+    def nlayers(self):
+        '''Return number of convolutional layers along the depth.
+        '''
+        if len(self.__layers) == 0:
+            return 0
+        n_layers = functools.reduce(lambda x, y: x + 2 * y, self.__layers[:-1], self.__layers[-1]) + 2
+        return n_layers
 
     def forward(self, x):
         x = self.conv_first(x)
@@ -314,6 +326,7 @@ def __init__(self, order, channel, layers, kernel_size=3, in_planes=1, out_plane
         super().__init__()
         if len(layers) < 2:
             raise ValueError('modules.conv: The argument "layers" should contain at least 2 values, but provide "{0}"'.format(layers))
+        self.__layers = layers
         ConvNd = get_convnd(order=order)
 
         ksize_e, psize_e, _ = cal_kernel_padding(kernel_size, ksize_plus=2)
@@ -345,7 +358,16 @@ def __init__(self, order, channel, layers, kernel_size=3, in_planes=1, out_plane
         self.conv_up_list.append(
             _BlockConvStkNd(order, channel, channel, hidden_planes=channel, kernel_size=ksize, padding=psize,
                             stride=1, stack_level=layers[0], scaler='down'))
-        self.conv_final = ConvNd(channel, in_planes, kernel_size=ksize_e, stride=1, padding=psize_e, bias=True)
+        self.conv_final = ConvNd(channel, out_planes, kernel_size=ksize_e, stride=1, padding=psize_e, bias=True)
+
+    @property
+    def nlayers(self):
+        '''Return number of convolutional layers along the depth.
+        '''
+        if len(self.__layers) == 0:
+            return 0
+        n_layers = functools.reduce(lambda x, y: x + 2 * y, self.__layers[:-1], self.__layers[-1]) + 2
+        return n_layers
 
     @staticmethod
     def cropping(x, x_ref_s):
@@ -399,6 +421,7 @@ def __init__(self, order, channel, layers, kernel_size=3, in_planes=1, out_lengt
         super().__init__()
         if len(layers) < 2:
             raise ValueError('modules.conv: The argument "layers" should contain at least 2 values, but provide "{0}"'.format(layers))
+        self.__layers = layers
         ConvNd = get_convnd(order=order)
 
         ksize_e, psize_e, _ = cal_kernel_padding(kernel_size, ksize_plus=2)
@@ -422,12 +445,23 @@ def __init__(self, order, channel, layers, kernel_size=3, in_planes=1, out_lengt
         if self.is_out_vector:
             self.fc = nn.Linear(channel, out_length, bias=True)
 
+    @property
+    def nlayers(self):
+        '''Return number of convolutional layers along the depth.
+        '''
+        if len(self.__layers) == 0:
+            return 0
+        n_layers = functools.reduce(lambda x, y: x + y, self.__layers, 0) + 2
+        return n_layers
+
     def forward(self, x):
         for layer in self.netbody:
             x = layer(x)
         if self.is_out_vector:
             x = torch.flatten(x, 1)
             return self.fc(x)
+        else:
+            return x
 
 
 class _DecoderNetNd(nn.Module):
@@ -462,12 +496,17 @@ def __init__(self, order, channel, layers, out_size, kernel_size=3, in_length=2,
         super().__init__()
         if len(layers) < 2:
             raise ValueError('modules.conv: The argument "layers" should contain at least 2 values, but provide "{0}"'.format(layers))
+        self.__layers = layers
+        self.__in_length = in_length
         ConvNd = get_convnd(order=order)
         ksize_e, psize_e, _ = cal_kernel_padding(kernel_size, ksize_plus=2)
         ksize, psize, stride = cal_kernel_padding(kernel_size)
         self.__order = order
+        if isinstance(out_size, int):
+            out_size = (out_size, ) * order
         self.shapes = cal_scaled_shapes(out_size, level=len(layers), stride=stride)
         channels = tuple(map(lambda n: channel * (2 ** n), range(len(layers) - 1, -1, -1)))
+        self.__in_channel = channels[0]
 
         # Require to convert the vector into channels
         self.is_in_vector = (in_length is not None and in_length > 0)
@@ -489,6 +528,22 @@ def __init__(self, order, channel, layers, out_size, kernel_size=3, in_length=2,
         self.netbody = netbody
         self.conv_last = ConvNd(channels[-1], out_planes, kernel_size=ksize_e, stride=1, padding=psize_e, bias=True)
 
+    @property
+    def input_size(self):
+        if self.is_in_vector:
+            return (self.__in_length, )
+        else:
+            return (self.__in_channel, *self.shapes[-1])
+
+    @property
+    def nlayers(self):
+        '''Return number of convolutional layers along the depth.
+        '''
+        if len(self.__layers) == 0:
+            return 0
+        n_layers = functools.reduce(lambda x, y: x + y, self.__layers, 0) + (3 if self.is_in_vector else 2)
+        return n_layers
+
     @staticmethod
     def cropping(x, x_ref_s):
         x_size = x.shape[2:]
@@ -672,9 +727,9 @@ def __init__(self, channel, layers, kernel_size=3, in_planes=1, out_planes=1):
                     of convolutional layers of a stage. The stage numer, i.e.
                     the depth of the network is the length of this list.
         Arguments (optional):
+            kernel_size: the kernel size of each block.
             in_planes: the channel number of the input data.
             out_planes: the channel number of the output data.
-            kernel_size: the kernel size of each block.
         '''
         super().__init__(1, channel=channel, layers=layers, kernel_size=kernel_size,
                          in_planes=in_planes, out_planes=out_planes)
@@ -698,9 +753,9 @@ def __init__(self, channel, layers, kernel_size=3, in_planes=1, out_planes=1):
                     of convolutional layers of a stage. The stage numer, i.e.
                     the depth of the network is the length of this list.
         Arguments (optional):
+            kernel_size: the kernel size of each block.
             in_planes: the channel number of the input data.
             out_planes: the channel number of the output data.
-            kernel_size: the kernel size of each block.
         '''
         super().__init__(order=2, channel=channel, layers=layers, kernel_size=kernel_size,
                          in_planes=in_planes, out_planes=out_planes)
@@ -724,9 +779,9 @@ def __init__(self, channel, layers, kernel_size=3, in_planes=1, out_planes=1):
                     of convolutional layers of a stage. The stage numer, i.e.
                     the depth of the network is the length of this list.
         Arguments (optional):
+            kernel_size: the kernel size of each block.
             in_planes: the channel number of the input data.
             out_planes: the channel number of the output data.
-            kernel_size: the kernel size of each block.
         '''
         super().__init__(order=3, channel=channel, layers=layers, kernel_size=kernel_size,
                          in_planes=in_planes, out_planes=out_planes)
@@ -1267,7 +1322,7 @@ def decnet16(out_size, order=2, **kwargs):
     '''Constructs a conv.DecoderNet-16 model.
     Configurations:
         Network depth: 5
-        Stage details: [2, 2, 3, 3, 3]
+        Stage details: [3, 3, 2, 2, 2]
         First channel number: 64
     Arguments:
         out_size: the output shape of the network.
@@ -1276,7 +1331,7 @@ def decnet16(out_size, order=2, **kwargs):
     Other Arguments (see mdnc.modules.conv.DecoderNet*d):
         in_length, out_planes, kernel_size
     '''
-    model = __get_decnet_nd(order)(64, [2, 2, 3, 3, 3], out_size=out_size, **kwargs)
+    model = __get_decnet_nd(order)(64, [3, 3, 2, 2, 2], out_size=out_size, **kwargs)
     return model
 
 
diff --git a/mdnc/modules/resnet.py b/mdnc/modules/resnet.py
index 97c20e1..4d4bfab 100644
--- a/mdnc/modules/resnet.py
+++ b/mdnc/modules/resnet.py
@@ -2,13 +2,13 @@
 # -*- coding: UTF-8 -*-
 '''
 ################################################################
-# Modules - 1D residual network
+# Modules - residual network
 # @ Modern Deep Network Toolkits for pyTorch
 # Yuchen Jin @ cainmagi@gmail.com
 # Requirements: (Pay attention to version)
 #   python 3.5+
 #   pyTorch 1.0.0+
-# This module is the definition of the 1D residual network. The
+# This module is the definition of the residual network. The
 # network could be initialized here and used for training and
 # processing.
 # The codes are inspired by:
@@ -16,6 +16,8 @@
 ################################################################
 '''
 
+import functools
+
 import torch
 import torch.nn as nn
 
@@ -28,6 +30,16 @@
            'DecoderNet1d', 'DecoderNet2d', 'DecoderNet3d', 'decnet13', 'decnet33', 'decnet48', 'decnet63']
 
 
+def get_block_depth(block):
+    block = block.casefold()
+    if block == 'bottleneck':
+        return 3
+    elif block == 'plain':
+        return 2
+    else:
+        raise ValueError('module.resnet: The argument "block" should be "plain" or "bottleneck".')
+
+
 class _BlockFactory:
     '''N-D block factory class for the residual network
     The block factory used for building any specialized block.
@@ -90,7 +102,7 @@ def res_branch(self, in_planes, out_planes, kernel_size=3, stride=1, padding=1,
             order, normalizer, activator, layer_order.
         '''
         ConvNd = get_convnd(order=self.order)
-        is_stride = check_is_stride(stride)
+        is_stride = check_is_stride(stride, output_size=output_size, scaler=scaler)
         seq = []
         if self.normalizer == 'null':
             if (not is_stride) or scaler == 'down':
@@ -162,7 +174,7 @@ def in_branch(self, in_planes, out_planes, stride=1, output_size=None, scaler='d
         Arguments (inherited):
             order, normalizer.
         '''
-        if check_is_stride(stride) or in_planes != out_planes:
+        if check_is_stride(stride, output_size=output_size, scaler=scaler) or in_planes != out_planes:
             ConvNd = get_convnd(order=self.order)
             seq = []
             if self.normalizer in 'null':
@@ -220,7 +232,7 @@ def __init__(self, order, in_planes, out_planes, kernel_size=3, stride=1, paddin
         Arguments (optional):
             kernel_size: the kernel size of this layer.
             stride: the stride size of this layer.
-            padding: the padding size of the convolutional layer.
+            padding: the padding size of the residual block.
             output_size: the size of the output data. This option is only used
                          when "scaler=up". When setting this value, the size
                          of the up-sampling would be given explicitly and
@@ -284,7 +296,7 @@ def __init__(self, order, in_planes, out_planes, kernel_size=3, stride=1, paddin
         Arguments (optional):
             kernel_size: the kernel size of this layer.
             stride: the stride size of this layer.
-            padding: the padding size of the convolutional layer.
+            padding: the padding size of the residual block.
             output_size: the size of the output data. This option is only used
                          when "scaler=up". When setting this value, the size
                          of the up-sampling would be given explicitly and
@@ -350,11 +362,11 @@ def __init__(self, order, in_planes, out_planes, block='bottleneck',
                            also used as the base of the following channels. If
                            not set, would use "out_planes" as the default
                            value.
-            kernel_size: the kernel size of the convolutional layers.
-            padding: the padding size of the convolutional layers.
-            stride: the stride size of the convolutional layers.
-            stack_level: the number of convolutional layers in this block,
-                         requiring to be >= 1.
+            kernel_size: the kernel size of the residual blocks.
+            padding: the padding size of the residual blocks.
+            stride: the stride size of the residual blocks.
+            stack_level: the number of residual blocks in this block, requiring
+                         to be >= 1.
             ex_planes: the channel number of the second input data. This value
                        is =0 in most time, but >0 during the decoding phase
                        of the U-Net. the extra input would be concatenated
@@ -388,7 +400,7 @@ def __init__(self, order, in_planes, out_planes, block='bottleneck',
                 Block(order, (in_planes + ex_planes) if i == 0 else hidden_planes, hidden_planes,
                       kernel_size=kernel_size, padding=padding, stride=1, scaler='down')
             )
-        self.conv_scale = Block(order, hidden_planes, out_planes,
+        self.conv_scale = Block(order, hidden_planes if stack_level > 1 else (in_planes + ex_planes), out_planes,
                                 kernel_size=kernel_size, padding=padding, stride=stride, scaler=scaler)
 
     @staticmethod
@@ -442,13 +454,15 @@ def __init__(self, order, channel, layers, block='bottleneck', kernel_size=3, in
         Arguments (optional):
             block: the block type, could be:
                    - bottleneck, - plain
+            kernel_size: the kernel size of each block.
             in_planes: the channel number of the input data.
             out_planes: the channel number of the output data.
-            kernel_size: the kernel size of each block.
         '''
         super().__init__()
         if len(layers) < 2:
             raise ValueError('modules.resnet: The argument "layers" should contain at least 2 values, but provide "{0}"'.format(layers))
+        self.__layers = layers
+        self.__block_depth = get_block_depth(block)
         ConvNd = get_convnd(order=order)
 
         ksize_e, psize_e, _ = cal_kernel_padding(kernel_size, ksize_plus=2)
@@ -480,7 +494,16 @@ def __init__(self, order, channel, layers, block='bottleneck', kernel_size=3, in
         self.conv_up_list.append(
             _BlockResStkNd(order, channel, channel, hidden_planes=channel, block=block,
                            kernel_size=ksize, padding=psize, stride=1, stack_level=layers[0], ex_planes=channel, scaler='down'))
-        self.conv_final = ConvNd(channel, in_planes, kernel_size=ksize_e, stride=1, padding=psize_e, bias=True)
+        self.conv_final = ConvNd(channel, out_planes, kernel_size=ksize_e, stride=1, padding=psize_e, bias=True)
+
+    @property
+    def nlayers(self):
+        '''Return number of convolutional layers along the depth.
+        '''
+        if len(self.__layers) == 0:
+            return 0
+        n_layers = functools.reduce(lambda x, y: x + 2 * self.__block_depth * y, self.__layers[:-1], self.__block_depth * self.__layers[-1]) + 2
+        return n_layers
 
     def forward(self, x):
         x = self.conv_first(x)
@@ -523,6 +546,8 @@ def __init__(self, order, channel, layers, block='bottleneck', kernel_size=3, in
         super().__init__()
         if len(layers) < 2:
             raise ValueError('modules.resnet: The argument "layers" should contain at least 2 values, but provide "{0}"'.format(layers))
+        self.__layers = layers
+        self.__block_depth = get_block_depth(block)
         ConvNd = get_convnd(order=order)
 
         ksize_e, psize_e, _ = cal_kernel_padding(kernel_size, ksize_plus=2)
@@ -554,7 +579,16 @@ def __init__(self, order, channel, layers, block='bottleneck', kernel_size=3, in
         self.conv_up_list.append(
             _BlockResStkNd(order, channel, channel, hidden_planes=channel, block=block,
                            kernel_size=ksize, padding=psize, stride=1, stack_level=layers[0], scaler='down'))
-        self.conv_final = ConvNd(channel, in_planes, kernel_size=ksize_e, stride=1, padding=psize_e, bias=True)
+        self.conv_final = ConvNd(channel, out_planes, kernel_size=ksize_e, stride=1, padding=psize_e, bias=True)
+
+    @property
+    def nlayers(self):
+        '''Return number of convolutional layers along the depth.
+        '''
+        if len(self.__layers) == 0:
+            return 0
+        n_layers = functools.reduce(lambda x, y: x + 2 * self.__block_depth * y, self.__layers[:-1], self.__block_depth * self.__layers[-1]) + 2
+        return n_layers
 
     @staticmethod
     def cropping(x, x_ref_s):
@@ -610,6 +644,8 @@ def __init__(self, order, channel, layers, block='bottleneck', kernel_size=3, in
         super().__init__()
         if len(layers) < 2:
             raise ValueError('modules.resnet: The argument "layers" should contain at least 2 values, but provide "{0}"'.format(layers))
+        self.__layers = layers
+        self.__block_depth = get_block_depth(block)
         ConvNd = get_convnd(order=order)
 
         ksize_e, psize_e, _ = cal_kernel_padding(kernel_size, ksize_plus=2)
@@ -633,12 +669,23 @@ def __init__(self, order, channel, layers, block='bottleneck', kernel_size=3, in
         if self.is_out_vector:
             self.fc = nn.Linear(channel, out_length, bias=True)
 
+    @property
+    def nlayers(self):
+        '''Return number of convolutional layers along the depth.
+        '''
+        if len(self.__layers) == 0:
+            return 0
+        n_layers = functools.reduce(lambda x, y: x + self.__block_depth * y, self.__layers, 0) + 2
+        return n_layers
+
     def forward(self, x):
         for layer in self.netbody:
             x = layer(x)
         if self.is_out_vector:
             x = torch.flatten(x, 1)
             return self.fc(x)
+        else:
+            return x
 
 
 class _DecoderNetNd(nn.Module):
@@ -675,12 +722,18 @@ def __init__(self, order, channel, layers, out_size, block='bottleneck', kernel_
         super().__init__()
         if len(layers) < 2:
             raise ValueError('modules.conv: The argument "layers" should contain at least 2 values, but provide "{0}"'.format(layers))
+        self.__layers = layers
+        self.__block_depth = get_block_depth(block)
+        self.__in_length = in_length
         ConvNd = get_convnd(order=order)
         ksize_e, psize_e, _ = cal_kernel_padding(kernel_size, ksize_plus=2)
         ksize, psize, stride = cal_kernel_padding(kernel_size)
         self.__order = order
+        if isinstance(out_size, int):
+            out_size = (out_size, ) * order
         self.shapes = cal_scaled_shapes(out_size, level=len(layers), stride=stride)
         channels = tuple(map(lambda n: channel * (2 ** n), range(len(layers) - 1, -1, -1)))
+        self.__in_channel = channels[0]
 
         # Require to convert the vector into channels
         self.is_in_vector = (in_length is not None and in_length > 0)
@@ -702,6 +755,22 @@ def __init__(self, order, channel, layers, out_size, block='bottleneck', kernel_
         self.netbody = netbody
         self.conv_last = ConvNd(channels[-1], out_planes, kernel_size=ksize_e, stride=1, padding=psize_e, bias=True)
 
+    @property
+    def input_size(self):
+        if self.is_in_vector:
+            return (self.__in_length, )
+        else:
+            return (self.__in_channel, *self.shapes[-1])
+
+    @property
+    def nlayers(self):
+        '''Return number of convolutional layers along the depth.
+        '''
+        if len(self.__layers) == 0:
+            return 0
+        n_layers = functools.reduce(lambda x, y: x + self.__block_depth * y, self.__layers, 0) + (3 if self.is_in_vector else 2)
+        return n_layers
+
     @staticmethod
     def cropping(x, x_ref_s):
         x_size = x.shape[2:]
@@ -746,7 +815,7 @@ def __init__(self, in_planes, out_planes, kernel_size=3, stride=1, padding=1, ou
         Arguments (optional):
             kernel_size: the kernel size of this layer.
             stride: the stride size of this layer.
-            padding: the padding size of the convolutional layer.
+            padding: the padding size of the residual block.
             output_size: the size of the output data. This option is only used
                          when "scaler=up". When setting this value, the size
                          of the up-sampling would be given explicitly and
@@ -794,7 +863,7 @@ def __init__(self, in_planes, out_planes, kernel_size=3, stride=1, padding=1, ou
         Arguments (optional):
             kernel_size: the kernel size of this layer.
             stride: the stride size of this layer.
-            padding: the padding size of the convolutional layer.
+            padding: the padding size of the residual block.
             output_size: the size of the output data. This option is only used
                          when "scaler=up". When setting this value, the size
                          of the up-sampling would be given explicitly and
@@ -842,7 +911,7 @@ def __init__(self, in_planes, out_planes, kernel_size=3, stride=1, padding=1, ou
         Arguments (optional):
             kernel_size: the kernel size of this layer.
             stride: the stride size of this layer.
-            padding: the padding size of the convolutional layer.
+            padding: the padding size of the residual block.
             output_size: the size of the output data. This option is only used
                          when "scaler=up". When setting this value, the size
                          of the up-sampling would be given explicitly and
@@ -890,7 +959,7 @@ def __init__(self, in_planes, out_planes, kernel_size=3, stride=1, padding=1, ou
         Arguments (optional):
             kernel_size: the kernel size of this layer.
             stride: the stride size of this layer.
-            padding: the padding size of the convolutional layer.
+            padding: the padding size of the residual block.
             output_size: the size of the output data. This option is only used
                          when "scaler=up". When setting this value, the size
                          of the up-sampling would be given explicitly and
@@ -986,7 +1055,7 @@ def __init__(self, in_planes, out_planes, kernel_size=3, stride=1, padding=1, ou
         Arguments (optional):
             kernel_size: the kernel size of this layer.
             stride: the stride size of this layer.
-            padding: the padding size of the convolutional layer.
+            padding: the padding size of the residual block.
             output_size: the size of the output data. This option is only used
                          when "scaler=up". When setting this value, the size
                          of the up-sampling would be given explicitly and
@@ -1034,9 +1103,9 @@ def __init__(self, channel, layers, block='bottleneck', kernel_size=3, in_planes
         Arguments (optional):
             block: the block type, could be:
                    - bottleneck, - plain
+            kernel_size: the kernel size of each block.
             in_planes: the channel number of the input data.
             out_planes: the channel number of the output data.
-            kernel_size: the kernel size of each block.
         '''
         super().__init__(1, channel=channel, layers=layers, block=block, kernel_size=kernel_size,
                          in_planes=in_planes, out_planes=out_planes)
@@ -1062,9 +1131,9 @@ def __init__(self, channel, layers, block='bottleneck', kernel_size=3, in_planes
         Arguments (optional):
             block: the block type, could be:
                    - bottleneck, - plain
+            kernel_size: the kernel size of each block.
             in_planes: the channel number of the input data.
             out_planes: the channel number of the output data.
-            kernel_size: the kernel size of each block.
         '''
         super().__init__(2, channel=channel, layers=layers, block=block, kernel_size=kernel_size,
                          in_planes=in_planes, out_planes=out_planes)
@@ -1090,9 +1159,9 @@ def __init__(self, channel, layers, block='bottleneck', kernel_size=3, in_planes
         Arguments (optional):
             block: the block type, could be:
                    - bottleneck, - plain
+            kernel_size: the kernel size of each block.
             in_planes: the channel number of the input data.
             out_planes: the channel number of the output data.
-            kernel_size: the kernel size of each block.
         '''
         super().__init__(3, channel=channel, layers=layers, block=block, kernel_size=kernel_size,
                          in_planes=in_planes, out_planes=out_planes)
diff --git a/mdnc/modules/utils.py b/mdnc/modules/utils.py
index 73fac77..0e9b059 100644
--- a/mdnc/modules/utils.py
+++ b/mdnc/modules/utils.py
@@ -80,7 +80,9 @@ def get_adaptive_pooling(order=2, out_size=1):
         raise ValueError('modules.utils: The argument "order" could only be 1, 2, or 3.')
 
 
-def check_is_stride(stride):
+def check_is_stride(stride, output_size=None, scaler='down'):
+    if output_size is not None and scaler == 'up':
+        return True
     if isinstance(stride, (list, tuple)):
         for s in stride:
             if s > 1: