caffe resnet152 to tf converts ok.

Author: tiqint

mmdnn/conversion/caffe/mapper.py

@@ -33,20 +33,21 @@ def _convert_output_shape(cls, kwargs, node):
         kwargs['_output_shapes'] = [shape]
 
     @classmethod
-    def get_kernel_params(cls, node):
+    def get_kernel_params(cls, node, input_shape):
         kwargs = {}
-        if node.kernel_parameters.p_h > 0 or node.kernel_parameters.p_w > 0:
-            padding = [0, node.kernel_parameters.p_h, node.kernel_parameters.p_w, 0, 0, node.kernel_parameters.p_h, node.kernel_parameters.p_w, 0]
-        elif node.kernel_parameters.s_h > 1 or node.kernel_parameters.s_w > 1:
-            padding = [0, (node.kernel_parameters.s_h - 1) // 2, (node.kernel_parameters.s_w - 1) // 2, 0, 0, node.kernel_parameters.s_h // 2, node.kernel_parameters.s_w // 2, 0]
-        else:
-            padding = None
-
-        kwargs['auto_pad'] = 'VALID'
+        
+        o_h_caffe = node.output_shape.height
+        o_h_tf = (input_shape.height + node.kernel_parameters.p_h * 2 - node.kernel_parameters.k_h + 1) // node.kernel_parameters.s_h
+        
+        o_w_caffe = node.output_shape.width
+        o_w_tf = (input_shape.width + node.kernel_parameters.p_w * 2 - node.kernel_parameters.k_w + 1) // node.kernel_parameters.s_w
+                
+        kwargs['pads'] = [0, node.kernel_parameters.p_h, node.kernel_parameters.p_w, 0] + \
+                  [0, node.kernel_parameters.p_h + o_h_caffe - o_h_tf, node.kernel_parameters.p_w + o_w_caffe - o_w_tf, 0]
         kwargs['strides'] = [1, node.kernel_parameters.s_h, node.kernel_parameters.s_w, 1]
         cls._convert_output_shape(kwargs, node)
 
-        return kwargs, {'pads' : padding, 'mode' : 'constant', 'constant_values' : 0.0}
+        return kwargs
 
 
     @classmethod
@@ -72,30 +73,22 @@ def map_input(cls, node):
 
     @classmethod
     def map_convolution(cls, node):
-        kwargs, padding = cls.get_kernel_params(node)
         parent, _ = node.get_only_parent()
+        kwargs = cls.get_kernel_params(node, parent.output_shape)
         kwargs['kernel_shape'] = [node.kernel_parameters.k_h, node.kernel_parameters.k_w, parent.output_shape.channels, node.parameters.num_output]
-        kwargs['use_bias'] = node.parameters.bias_term
-        group = node.parameters.group
-        if group != 1:
-            kwargs['group'] = group
-
-        if padding['pads'] != None:
-            return [Node.create('Pad', **padding), Node.create('Conv', **kwargs)]
-        else:
-            kwargs['pads'] = [0] * 8
-            return Node.create('Conv', **kwargs)
+        kwargs['use_bias'] = node.parameters.bias_term        
+        kwargs['group'] = node.parameters.group
+        return Node.create('Conv', **kwargs)
 
 
     @classmethod
     def map_deconvolution(cls, node):
         raise NotImplementedError()
-        kwargs = cls.get_kernel_params(node)
         parent, _ = node.get_only_parent()
-        kwargs['kernel_shape'] = [node.kernel_parameters.k_h, node.kernel_parameters.k_w, parent.output_shape.channels, node.parameters.num_output]
-        group = node.parameters.group
-        if group != 1:
-            kwargs['group'] = group
+        kwargs = cls.get_kernel_params(node, parent.output_shape)
+        
+        kwargs['kernel_shape'] = [node.kernel_parameters.k_h, node.kernel_parameters.k_w, parent.output_shape.channels, node.parameters.num_output]        
+        kwargs['group'] = node.parameters.group
         return Node.create('deconv', **kwargs)
 
     @classmethod
@@ -115,7 +108,8 @@ def map_relu(cls, node):
 
     @classmethod
     def map_pooling(cls, node):
-        kwargs, padding = cls.get_kernel_params(node)
+        parent, _ = node.get_only_parent()
+        kwargs = cls.get_kernel_params(node, parent.output_shape)
         if node.parameters.pool == 0:
             kwargs['pooling_type'] = 'MAX'
         elif node.parameters.pool == 1:
@@ -125,12 +119,7 @@ def map_pooling(cls, node):
             raise ConversionError('Unsupported pooling type.')
         kwargs['kernel_shape'] = [1, node.kernel_parameters.k_h, node.kernel_parameters.k_w, 1]
         cls._convert_output_shape(kwargs, node)
-
-        if padding['pads'] != None:
-            return [Node.create('Pad', **padding), Node.create('Pool', **kwargs)]
-        else:
-            kwargs['pads'] = [0] * 8
-            return Node.create('Pool', **kwargs)
+        return Node.create('Pool', **kwargs)
 
 
     @classmethod

mmdnn/conversion/examples/tensorflow/imagenet_test.py

@@ -15,7 +15,7 @@ class TestTF(TestKit):
     def __init__(self):
         super(TestTF, self).__init__()
         self.input, self.model = self.MainModel.KitModel(self.args.w)
-        # self.input, self.model, self.testop = KitModel(os.path.abspath('.') + '/kit_imagenet.npy')
+        # self.input, self.model, self.testop = self.MainModel.KitModel(self.args.w)
 
 
     def preprocess(self, image_path):
@@ -24,7 +24,9 @@ def preprocess(self, image_path):
 
 
     def print_result(self):
-        with tf.Session() as sess:            
+        with tf.Session() as sess:
+            writer = tf.summary.FileWriter('./graphs', sess.graph)
+            writer.close()
             init = tf.global_variables_initializer()
             sess.run(init)
             predict = sess.run(self.model, feed_dict = {self.input : self.data})
@@ -33,20 +35,20 @@ def print_result(self):
 
 
     def print_intermediate_result(self, layer_name, if_transpose = False):
-        testop = tf.get_default_graph().get_operation_by_name(layer_name)
-        # testop = self.testop
+        # testop = tf.get_default_graph().get_operation_by_name(layer_name)
+        testop = self.testop
         with tf.Session() as sess:
             init = tf.global_variables_initializer()
             sess.run(init)
             intermediate_output = sess.run(testop, feed_dict = {self.input : self.data})
 
-        super(TestTF, self).predict(intermediate_output, if_transpose)
+        super(TestTF, self).print_intermediate_result(intermediate_output, if_transpose)
 
 
     def inference(self, image_path):
         self.preprocess(image_path)
 
-        # self.print_intermediate_result('conv1_7x7_s2_1', False)
+        # self.print_intermediate_result('conv1_7x7_s2_1', True)
 
         self.print_result()