diff --git a/mmdnn/conversion/tensorflow/README.md b/mmdnn/conversion/tensorflow/README.md
index c1ddd85c..4f2016cd 100644
--- a/mmdnn/conversion/tensorflow/README.md
+++ b/mmdnn/conversion/tensorflow/README.md
@@ -12,6 +12,7 @@ We tested the [slim pre-trained models](https://github.com/tensorflow/models/tre
 |   ResNet V1  |   x   |    √   |   o  |   √   |   √   |    √    |      √     |   √  |
 |   ResNet V2  |   x   |    √   |   √  |   √   |   √   |    √    |      √     |   √  |
 | MobileNet V1 |   x   |    √   |   o  |   √   |   √   |    √    |      √     |   √  |
+| MobileNet V2 |   x   |    √   |   o  |   √   |   √   |    √    |      √     |   √  |
 |   NasNet-A   |   x   |        |      |       |   √   |    √    |      √     |   √  |
 
 **√** - Correctness tested
diff --git a/setup.py b/setup.py
index fee6f0f8..a5b32496 100644
--- a/setup.py
+++ b/setup.py
@@ -75,7 +75,8 @@
         'numpy >= 1.11.0',
         'protobuf >= 3.1.0',
         'six >= 1.10.0',
-        'uuid'
+        'uuid',
+        'pillow >= 3.1.0',
     ],
 
     # To provide executable scripts, use entry points in preference to the
diff --git a/tests/test_conversion_imagenet.py b/tests/test_conversion_imagenet.py
index b7ed3dbb..1936276f 100644
--- a/tests/test_conversion_imagenet.py
+++ b/tests/test_conversion_imagenet.py
@@ -703,6 +703,7 @@ def OnnxEmit(original_framework, architecture_name, architecture_path, weight_pa
         'tensorflow_Cntk_resnet_v1_152',            # Cntk Padding is SAME_LOWER, but Tensorflow Padding is SAME_UPPER, in first convolution layer.
         'tensorflow_Cntk_resnet_v2_152',            # Cntk Padding is SAME_LOWER, but Tensorflow Padding is SAME_UPPER, in first convolution layer.
         'tensorflow_Cntk_mobilenet_v1_1.0',         # Cntk Padding is SAME_LOWER, but Tensorflow Padding is SAME_UPPER, in first convolution layer.
+        'tensorflow_Cntk_mobilenet_v2_1.0_224',     # Cntk Padding is SAME_LOWER, but Tensorflow Padding is SAME_UPPER, in first convolution layer.
         'tensorflow_frozen_MXNet_inception_v1',     # different after AvgPool. AVG POOL padding difference between these two framework. MXNet AVGPooling Padding is SAME_LOWER, Tensorflow AVGPooling Padding is SAME_UPPER
         'tensorflow_MXNet_inception_v3',            # different after "InceptionV3/InceptionV3/Mixed_5b/Branch_3/AvgPool_0a_3x3/AvgPool". AVG POOL padding difference between these two framework.
         'caffe_Pytorch_inception_v1',               # TODO
@@ -766,7 +767,7 @@ def OnnxEmit(original_framework, architecture_name, architecture_path, weight_pa
             'resnet_v2_50'          : [CaffeEmit, CoreMLEmit, KerasEmit, MXNetEmit, OnnxEmit, PytorchEmit, TensorflowEmit], # TODO: CntkEmit
             'resnet_v2_152'         : [CaffeEmit, CoreMLEmit, CntkEmit, KerasEmit, MXNetEmit, OnnxEmit, PytorchEmit, TensorflowEmit],
             'mobilenet_v1_1.0'      : [CoreMLEmit, CntkEmit, KerasEmit, MXNetEmit, OnnxEmit, PytorchEmit, TensorflowEmit], # TODO: CaffeEmit(Crash)
-            'mobilenet_v2_1.0_224'  : [CoreMLEmit, KerasEmit, MXNetEmit, OnnxEmit, PytorchEmit, TensorflowEmit], # TODO: CaffeEmit(Crash) CntkEmit
+            'mobilenet_v2_1.0_224'  : [CoreMLEmit, CntkEmit, KerasEmit, MXNetEmit, OnnxEmit, PytorchEmit, TensorflowEmit], # TODO: CaffeEmit(Crash)
             'nasnet-a_large'        : [MXNetEmit, OnnxEmit, PytorchEmit, TensorflowEmit], # TODO: KerasEmit(Slice Layer: https://blog.csdn.net/lujiandong1/article/details/54936185)
             # 'inception_resnet_v2' : [TensorflowEmit], # TODO PytorchEmit