Merge pull request #555 from will-am/xception

Add Xception model for image classification

README.cn.md

@@ -98,7 +98,7 @@ PaddlePaddle提供了丰富的运算单元，帮助大家以模块化的方式
 
 图像相比文字能够提供更加生动、容易理解及更具艺术感的信息，是人们转递与交换信息的重要来源。图像分类是根据图像的语义信息对不同类别图像进行区分，是计算机视觉中重要的基础问题，也是图像检测、图像分割、物体跟踪、行为分析等其他高层视觉任务的基础，在许多领域都有着广泛的应用。如：安防领域的人脸识别和智能视频分析等，交通领域的交通场景识别，互联网领域基于内容的图像检索和相册自动归类，医学领域的图像识别等。
 
-在图像分类任务中，我们向大家介绍如何训练AlexNet、VGG、GoogLeNet、ResNet、Inception-v4和Inception-Resnet-V2模型。同时提供了能够将Caffe或TensorFlow训练好的模型文件转换为PaddlePaddle模型文件的模型转换工具。
+在图像分类任务中，我们向大家介绍如何训练AlexNet、VGG、GoogLeNet、ResNet、Inception-v4、Inception-Resnet-V2和Xception模型。同时提供了能够将Caffe或TensorFlow训练好的模型文件转换为PaddlePaddle模型文件的模型转换工具。
 
 - 11.1 [将Caffe模型文件转换为PaddlePaddle模型文件](https://github.com/PaddlePaddle/models/tree/develop/image_classification/caffe2paddle)
 - 11.2 [将TensorFlow模型文件转换为PaddlePaddle模型文件](https://github.com/PaddlePaddle/models/tree/develop/image_classification/tf2paddle)
@@ -107,6 +107,7 @@ PaddlePaddle提供了丰富的运算单元，帮助大家以模块化的方式
 - 11.5 [Residual Network](https://github.com/PaddlePaddle/models/tree/develop/image_classification)
 - 11.6 [Inception-v4](https://github.com/PaddlePaddle/models/tree/develop/image_classification)
 - 11.7 [Inception-Resnet-V2](https://github.com/PaddlePaddle/models/tree/develop/image_classification)
+- 11.8 [Xception](https://github.com/PaddlePaddle/models/tree/develop/image_classification)
 
 ## 12. 目标检测
 

README.md

@@ -72,7 +72,7 @@ As an example for sequence-to-sequence learning, we take the machine translation
 
 ## 9. Image classification
 
-For the example of image classification, we show you how to train AlexNet, VGG, GoogLeNet, ResNet, Inception-v4 and Inception-Resnet-V2 models in PaddlePaddle. It also provides model conversion tools that convert Caffe or TensorFlow trained model files into PaddlePaddle model files.
+For the example of image classification, we show you how to train AlexNet, VGG, GoogLeNet, ResNet, Inception-v4, Inception-Resnet-V2 and Xception models in PaddlePaddle. It also provides model conversion tools that convert Caffe or TensorFlow trained model files into PaddlePaddle model files.
 
 - 9.1 [convert Caffe model file to PaddlePaddle model file](https://github.com/PaddlePaddle/models/tree/develop/image_classification/caffe2paddle)
 - 9.2 [convert TensorFlow model file to PaddlePaddle model file](https://github.com/PaddlePaddle/models/tree/develop/image_classification/tf2paddle)
@@ -81,5 +81,6 @@ For the example of image classification, we show you how to train AlexNet, VGG,
 - 9.5 [Residual Network](https://github.com/PaddlePaddle/models/tree/develop/image_classification)
 - 9.6 [Inception-v4](https://github.com/PaddlePaddle/models/tree/develop/image_classification)
 - 9.7 [Inception-Resnet-V2](https://github.com/PaddlePaddle/models/tree/develop/image_classification)
+- 9.8 [Xception](https://github.com/PaddlePaddle/models/tree/develop/image_classification)
 
 This tutorial is contributed by [PaddlePaddle](https://github.com/PaddlePaddle/Paddle) and licensed under the [Apache-2.0 license](LICENSE).

image_classification/README.md

 图像分类
 =======================
 
-这里将介绍如何在PaddlePaddle下使用AlexNet、VGG、GoogLeNet、ResNet、Inception-v4和Inception-ResNet-v2模型进行图像分类。图像分类问题的描述和这些模型的介绍可以参考[PaddlePaddle book](https://github.com/PaddlePaddle/book/tree/develop/03.image_classification)。
+这里将介绍如何在PaddlePaddle下使用AlexNet、VGG、GoogLeNet、ResNet、Inception-v4、Inception-ResNet-v2和Xception模型进行图像分类。图像分类问题的描述和这些模型的介绍可以参考[PaddlePaddle book](https://github.com/PaddlePaddle/book/tree/develop/03.image_classification)。
 
 ## 训练模型
 
@@ -22,6 +22,7 @@ import alexnet
 import googlenet
 import inception_v4
 import inception_resnet_v2
+import xception
 
 
 # PaddlePaddle init
@@ -47,7 +48,7 @@ lbl = paddle.layer.data(
 
 ### 获得所用模型
 
-这里可以选择使用AlexNet、VGG、GoogLeNet、ResNet、Inception-v4和Inception-ResNet-v2模型中的一个模型进行图像分类。通过调用相应的方法可以获得网络最后的Softmax层。
+这里可以选择使用AlexNet、VGG、GoogLeNet、ResNet、Inception-v4、Inception-ResNet-v2和Xception模型中的一个模型进行图像分类。通过调用相应的方法可以获得网络最后的Softmax层。
 
 1. 使用AlexNet模型
 
@@ -112,6 +113,14 @@ out = inception_resnet_v2.inception_resnet_v2(
 
 注意，由于和其他几种模型输入大小不同，若配合提供的`reader.py`使用Inception-ResNet-v2时请先将`reader.py`中`paddle.image.simple_transform`中的参数为修改为相应大小。
 
+7. 使用Xception模型
+
+Xception模型可以通过下面的代码获取：
+
+```python
+out = xception.xception(image, class_dim=CLASS_DIM)
+```
+
 ### 定义损失函数
 
 ```python
@@ -199,7 +208,7 @@ def event_handler(event):
 
 ### 定义训练方法
 
-对于AlexNet、VGG、ResNet、Inception-v4和Inception-ResNet-v2，可以按下面的代码定义训练方法：
+对于AlexNet、VGG、ResNet、Inception-v4、Inception-ResNet-v2和Xception，可以按下面的代码定义训练方法：
 
 ```python
 # Create trainer

image_classification/infer.py

@@ -12,6 +12,7 @@ import alexnet
 import googlenet
 import inception_v4
 import inception_resnet_v2
+import xception
 
 DATA_DIM = 3 * 224 * 224  # Use 3 * 331 * 331 or 3 * 299 * 299 for Inception-ResNet-v2.
 CLASS_DIM = 102
@@ -29,7 +30,7 @@ def main():
         help='The model for image classification',
         choices=[
             'alexnet', 'vgg13', 'vgg16', 'vgg19', 'resnet', 'googlenet',
-            'inception-resnet-v2', 'inception_v4'
+            'inception-resnet-v2', 'inception_v4', 'xception'
         ])
     parser.add_argument(
         'params_path', help='The file which stores the parameters')
@@ -59,6 +60,8 @@ def main():
             image, class_dim=CLASS_DIM, dropout_rate=0.5, data_dim=DATA_DIM)
     elif args.model == 'inception_v4':
         out = inception_v4.inception_v4(image, class_dim=CLASS_DIM)
+    elif args.model == 'xception':
+        out = xception.xception(image, class_dim=CLASS_DIM)
 
     # load parameters
     with gzip.open(args.params_path, 'r') as f:

image_classification/train.py

@@ -10,6 +10,7 @@ import alexnet
 import googlenet
 import inception_v4
 import inception_resnet_v2
+import xception
 
 DATA_DIM = 3 * 224 * 224  # Use 3 * 331 * 331 or 3 * 299 * 299 for Inception-ResNet-v2.
 CLASS_DIM = 102
@@ -24,7 +25,7 @@ def main():
         help='The model for image classification',
         choices=[
             'alexnet', 'vgg13', 'vgg16', 'vgg19', 'resnet', 'googlenet',
-            'inception-resnet-v2', 'inception_v4'
+            'inception-resnet-v2', 'inception_v4', 'xception'
         ])
     args = parser.parse_args()
 
@@ -64,6 +65,8 @@ def main():
             image, class_dim=CLASS_DIM, dropout_rate=0.5, data_dim=DATA_DIM)
     elif args.model == 'inception_v4':
         out = inception_v4.inception_v4(image, class_dim=CLASS_DIM)
+    elif args.model == 'xception':
+        out = xception.xception(image, class_dim=CLASS_DIM)
 
     cost = paddle.layer.classification_cost(input=out, label=lbl)
 

image_classification/xception.py

+import paddle.v2 as paddle
+
+__all__ = ['xception']
+
+
+def img_separable_conv_bn(name, input, num_channels, num_out_channels,
+                          filter_size, stride, padding, act):
+    conv = paddle.networks.img_separable_conv(
+        name=name,
+        input=input,
+        num_channels=num_channels,
+        num_out_channels=num_out_channels,
+        filter_size=filter_size,
+        stride=stride,
+        padding=padding,
+        act=paddle.activation.Linear())
+    norm = paddle.layer.batch_norm(name=name + '_norm', input=conv, act=act)
+    return norm
+
+
+def img_conv_bn(name, input, num_channels, num_filters, filter_size, stride,
+                padding, act):
+    conv = paddle.layer.img_conv(
+        name=name,
+        input=input,
+        num_channels=num_channels,
+        num_filters=num_filters,
+        filter_size=filter_size,
+        stride=stride,
+        padding=padding,
+        act=paddle.activation.Linear())
+    norm = paddle.layer.batch_norm(name=name + '_norm', input=conv, act=act)
+    return norm
+
+
+def conv_block0(input,
+                group,
+                num_channels,
+                num_filters,
+                num_filters2=None,
+                filter_size=3,
+                pool_padding=0,
+                entry_relu=True):
+    if num_filters2 is None:
+        num_filters2 = num_filters
+
+    if entry_relu:
+        act_input = paddle.layer.mixed(
+            input=paddle.layer.identity_projection(input=input),
+            act=paddle.activation.Relu())
+    else:
+        act_input = input
+    conv0 = img_separable_conv_bn(
+        name='xception_block{0}_conv0'.format(group),
+        input=act_input,
+        num_channels=num_channels,
+        num_out_channels=num_filters,
+        filter_size=filter_size,
+        stride=1,
+        padding=(filter_size - 1) / 2,
+        act=paddle.activation.Relu())
+    conv1 = img_separable_conv_bn(
+        name='xception_block{0}_conv1'.format(group),
+        input=conv0,
+        num_channels=num_filters,
+        num_out_channels=num_filters2,
+        filter_size=filter_size,
+        stride=1,
+        padding=(filter_size - 1) / 2,
+        act=paddle.activation.Linear())
+    pool0 = paddle.layer.img_pool(
+        name='xception_block{0}_pool'.format(group),
+        input=conv1,
+        pool_size=3,
+        stride=2,
+        padding=pool_padding,
+        num_channels=num_filters2,
+        pool_type=paddle.pooling.CudnnMax())
+
+    shortcut = img_conv_bn(
+        name='xception_block{0}_shortcut'.format(group),
+        input=input,
+        num_channels=num_channels,
+        num_filters=num_filters2,
+        filter_size=1,
+        stride=2,
+        padding=0,
+        act=paddle.activation.Linear())
+
+    return paddle.layer.addto(
+        input=[pool0, shortcut], act=paddle.activation.Linear())
+
+
+def conv_block1(input, group, num_channels, num_filters, filter_size=3):
+    act_input = paddle.layer.mixed(
+        input=paddle.layer.identity_projection(input=input),
+        act=paddle.activation.Relu())
+    conv0 = img_separable_conv_bn(
+        name='xception_block{0}_conv0'.format(group),
+        input=act_input,
+        num_channels=num_channels,
+        num_out_channels=num_filters,
+        filter_size=filter_size,
+        stride=1,
+        padding=(filter_size - 1) / 2,
+        act=paddle.activation.Relu())
+    conv1 = img_separable_conv_bn(
+        name='xception_block{0}_conv1'.format(group),
+        input=conv0,
+        num_channels=num_filters,
+        num_out_channels=num_filters,
+        filter_size=filter_size,
+        stride=1,
+        padding=(filter_size - 1) / 2,
+        act=paddle.activation.Relu())
+    conv2 = img_separable_conv_bn(
+        name='xception_block{0}_conv2'.format(group),
+        input=conv1,
+        num_channels=num_filters,
+        num_out_channels=num_filters,
+        filter_size=filter_size,
+        stride=1,
+        padding=(filter_size - 1) / 2,
+        act=paddle.activation.Linear())
+
+    shortcut = input
+    return paddle.layer.addto(
+        input=[conv2, shortcut], act=paddle.activation.Linear())
+
+
+def xception(input, class_dim):
+    conv = img_conv_bn(
+        name='xception_conv0',
+        input=input,
+        num_channels=3,
+        num_filters=32,
+        filter_size=3,
+        stride=2,
+        padding=1,
+        act=paddle.activation.Relu())
+    conv = img_conv_bn(
+        name='xception_conv1',
+        input=conv,
+        num_channels=32,
+        num_filters=64,
+        filter_size=3,
+        stride=1,
+        padding=1,
+        act=paddle.activation.Relu())
+    conv = conv_block0(
+        input=conv, group=2, num_channels=64, num_filters=128, entry_relu=False)
+    conv = conv_block0(input=conv, group=3, num_channels=128, num_filters=256)
+    conv = conv_block0(input=conv, group=4, num_channels=256, num_filters=728)
+    for group in range(5, 13):
+        conv = conv_block1(
+            input=conv, group=group, num_channels=728, num_filters=728)
+    conv = conv_block0(
+        input=conv,
+        group=13,
+        num_channels=728,
+        num_filters=728,
+        num_filters2=1024)
+    conv = img_separable_conv_bn(
+        name='xception_conv14',
+        input=conv,
+        num_channels=1024,
+        num_out_channels=1536,
+        filter_size=3,
+        stride=1,
+        padding=1,
+        act=paddle.activation.Relu())
+    conv = img_separable_conv_bn(
+        name='xception_conv15',
+        input=conv,
+        num_channels=1536,
+        num_out_channels=2048,
+        filter_size=3,
+        stride=1,
+        padding=1,
+        act=paddle.activation.Relu())
+    pool = paddle.layer.img_pool(
+        name='xception_global_pool',
+        input=conv,
+        pool_size=7,
+        stride=1,
+        num_channels=2048,
+        pool_type=paddle.pooling.CudnnAvg())
+    out = paddle.layer.fc(
+        name='xception_fc',
+        input=pool,
+        size=class_dim,
+        act=paddle.activation.Softmax())
+    return out