diff --git a/03.image_classification/README.en.md b/03.image_classification/README.en.md
index 2c243c62737cb217ec00df7103ac9b7b378e08d5..d5e8b650199cc32bec817a884baed222d17115e3 100644
--- a/03.image_classification/README.en.md
+++ b/03.image_classification/README.en.md
@@ -169,6 +169,7 @@ We must import and initialize PaddlePaddle (enable/disable GPU, set the number o
 
 ```python
 import sys
+import gzip
 import paddle.v2 as paddle
 from vgg import vgg_bn_drop
 from resnet import resnet_cifar10
@@ -437,6 +438,10 @@ def event_handler(event):
             sys.stdout.write('.')
             sys.stdout.flush()
     if isinstance(event, paddle.event.EndPass):
+        # save parameters
+        with gzip.open('params_pass_%d.tar.gz' % event.pass_id, 'w') as f:
+            parameters.to_tar(f)
+
         result = trainer.test(
             reader=paddle.batch(
                 paddle.dataset.cifar.test10(), batch_size=128),
@@ -475,7 +480,29 @@ Figure 12. The error rate of VGG model on CIFAR10
 </p>
 
 
-After training is done, the model from each pass is saved in `output/pass-%05d`. For example, the model of Pass 300 is saved in `output/pass-00299`.
+
+## Application
+
+After training is done, users can use the trained model to classify images. The following code shows how to infer through `paddle.infer` interface.
+
+```python
+from PIL import Image
+import numpy as np
+def load_image(file):
+    im = Image.open(file)
+    im = im.resize((32, 32), Image.ANTIALIAS)
+    im = np.array(im).astype(np.float32).flatten()
+    im = im / 255.0
+    return im
+test_data = []
+test_data.append((load_image('image/dog.png'),))
+
+probs = paddle.infer(
+    output_layer=out, parameters=parameters, input=test_data)
+lab = np.argsort(-probs) # probs and lab are the results of one batch data
+print "Label of image/dog.png is: %d" % lab[0][0]
+```
+
 
 ## Conclusion
 
diff --git a/03.image_classification/README.md b/03.image_classification/README.md
index 0eea28f906e1a63c1703c7d660e4bd9987df2a04..8f3a8d256173839adeec06125527a6b696b79974 100644
--- a/03.image_classification/README.md
+++ b/03.image_classification/README.md
@@ -156,6 +156,7 @@ Paddle API提供了自动加载cifar数据集模块 `paddle.dataset.cifar`。
 
 ```python
 import sys
+import gzip
 import paddle.v2 as paddle
 from vgg import vgg_bn_drop
 from resnet import resnet_cifar10
@@ -409,6 +410,7 @@ def event_handler_plot(event):
             cost_ploter.plot()
         step += 1
     if isinstance(event, paddle.event.EndPass):
+
         result = trainer.test(
             reader=paddle.batch(
                 paddle.dataset.cifar.test10(), batch_size=128),
@@ -429,6 +431,10 @@ def event_handler(event):
             sys.stdout.write('.')
             sys.stdout.flush()
     if isinstance(event, paddle.event.EndPass):
+        # save parameters
+        with gzip.open('params_pass_%d.tar.gz' % event.pass_id, 'w') as f:
+            parameters.to_tar(f)
+
         result = trainer.test(
             reader=paddle.batch(
                 paddle.dataset.cifar.test10(), batch_size=128),
@@ -467,6 +473,28 @@ Test with Pass 0, {'classification_error_evaluator': 0.885200023651123}
 图12. CIFAR10数据集上VGG模型的分类错误率
 </p>
 
+## 应用模型
+
+可以使用训练好的模型对图片进行分类，下面程序展示了如何使用`paddle.infer`接口进行推断。
+
+```python
+from PIL import Image
+import numpy as np
+def load_image(file):
+    im = Image.open(file)
+    im = im.resize((32, 32), Image.ANTIALIAS)
+    im = np.array(im).astype(np.float32).flatten()
+    im = im / 255.0
+    return im
+test_data = []
+test_data.append((load_image('image/dog.png'),))
+
+probs = paddle.infer(
+    output_layer=out, parameters=parameters, input=test_data)
+lab = np.argsort(-probs) # probs and lab are the results of one batch data
+print "Label of image/dog.png is: %d" % lab[0][0]
+```
+
 
 ## 总结
 
diff --git a/03.image_classification/index.en.html b/03.image_classification/index.en.html
index c28836ac766f49924fa10b56045de17178cde3e1..37f1dc5f2392a931f0e5161926abe262f3fad7ee 100644
--- a/03.image_classification/index.en.html
+++ b/03.image_classification/index.en.html
@@ -211,6 +211,7 @@ We must import and initialize PaddlePaddle (enable/disable GPU, set the number o
 
 ```python
 import sys
+import gzip
 import paddle.v2 as paddle
 from vgg import vgg_bn_drop
 from resnet import resnet_cifar10
@@ -479,6 +480,10 @@ def event_handler(event):
             sys.stdout.write('.')
             sys.stdout.flush()
     if isinstance(event, paddle.event.EndPass):
+        # save parameters
+        with gzip.open('params_pass_%d.tar.gz' % event.pass_id, 'w') as f:
+            parameters.to_tar(f)
+
         result = trainer.test(
             reader=paddle.batch(
                 paddle.dataset.cifar.test10(), batch_size=128),
@@ -517,7 +522,29 @@ Figure 12. The error rate of VGG model on CIFAR10
 </p>
 
 
-After training is done, the model from each pass is saved in `output/pass-%05d`. For example, the model of Pass 300 is saved in `output/pass-00299`.
+
+## Application
+
+After training is done, users can use the trained model to classify images. The following code shows how to infer through `paddle.infer` interface.
+
+```python
+from PIL import Image
+import numpy as np
+def load_image(file):
+    im = Image.open(file)
+    im = im.resize((32, 32), Image.ANTIALIAS)
+    im = np.array(im).astype(np.float32).flatten()
+    im = im / 255.0
+    return im
+test_data = []
+test_data.append((load_image('image/dog.png'),))
+
+probs = paddle.infer(
+    output_layer=out, parameters=parameters, input=test_data)
+lab = np.argsort(-probs) # probs and lab are the results of one batch data
+print "Label of image/dog.png is: %d" % lab[0][0]
+```
+
 
 ## Conclusion
 
diff --git a/03.image_classification/index.html b/03.image_classification/index.html
index 8678bde12918bbf847f810242041bd130be5cacb..4001d122d275657a201f47e70fb4923688e523e9 100644
--- a/03.image_classification/index.html
+++ b/03.image_classification/index.html
@@ -198,6 +198,7 @@ Paddle API提供了自动加载cifar数据集模块 `paddle.dataset.cifar`。
 
 ```python
 import sys
+import gzip
 import paddle.v2 as paddle
 from vgg import vgg_bn_drop
 from resnet import resnet_cifar10
@@ -451,6 +452,7 @@ def event_handler_plot(event):
             cost_ploter.plot()
         step += 1
     if isinstance(event, paddle.event.EndPass):
+
         result = trainer.test(
             reader=paddle.batch(
                 paddle.dataset.cifar.test10(), batch_size=128),
@@ -471,6 +473,10 @@ def event_handler(event):
             sys.stdout.write('.')
             sys.stdout.flush()
     if isinstance(event, paddle.event.EndPass):
+        # save parameters
+        with gzip.open('params_pass_%d.tar.gz' % event.pass_id, 'w') as f:
+            parameters.to_tar(f)
+
         result = trainer.test(
             reader=paddle.batch(
                 paddle.dataset.cifar.test10(), batch_size=128),
@@ -509,6 +515,28 @@ Test with Pass 0, {'classification_error_evaluator': 0.885200023651123}
 图12. CIFAR10数据集上VGG模型的分类错误率
 </p>
 
+## 应用模型
+
+可以使用训练好的模型对图片进行分类，下面程序展示了如何使用`paddle.infer`接口进行推断。
+
+```python
+from PIL import Image
+import numpy as np
+def load_image(file):
+    im = Image.open(file)
+    im = im.resize((32, 32), Image.ANTIALIAS)
+    im = np.array(im).astype(np.float32).flatten()
+    im = im / 255.0
+    return im
+test_data = []
+test_data.append((load_image('image/dog.png'),))
+
+probs = paddle.infer(
+    output_layer=out, parameters=parameters, input=test_data)
+lab = np.argsort(-probs) # probs and lab are the results of one batch data
+print "Label of image/dog.png is: %d" % lab[0][0]
+```
+
 
 ## 总结
 
diff --git a/03.image_classification/train.py b/03.image_classification/train.py
index f8e18452cc08f174ef3c63026c880759734e42f4..e60b9bda3c7ec1eefb2cfb9798f72aa8fd5e2b93 100644
--- a/03.image_classification/train.py
+++ b/03.image_classification/train.py
@@ -13,6 +13,7 @@
 # limitations under the License
 
 import sys
+import gzip
 
 import paddle.v2 as paddle
 
@@ -66,6 +67,10 @@ def main():
                 sys.stdout.write('.')
                 sys.stdout.flush()
         if isinstance(event, paddle.event.EndPass):
+            # save parameters
+            with gzip.open('params_pass_%d.tar.gz' % event.pass_id, 'w') as f:
+                parameters.to_tar(f)
+
             result = trainer.test(
                 reader=paddle.batch(
                     paddle.dataset.cifar.test10(), batch_size=128),
@@ -81,11 +86,30 @@ def main():
             paddle.reader.shuffle(
                 paddle.dataset.cifar.train10(), buf_size=50000),
             batch_size=128),
-        num_passes=200,
+        num_passes=1,
         event_handler=event_handler,
         feeding={'image': 0,
                  'label': 1})
 
+    # inference
+    from PIL import Image
+    import numpy as np
+
+    def load_image(file):
+        im = Image.open(file)
+        im = im.resize((32, 32), Image.ANTIALIAS)
+        im = np.array(im).astype(np.float32).flatten()
+        im = im / 255.0
+        return im
+
+    test_data = []
+    test_data.append((load_image('image/dog.png'), ))
+
+    probs = paddle.infer(
+        output_layer=out, parameters=parameters, input=test_data)
+    lab = np.argsort(-probs)  # probs and lab are the results of one batch data
+    print "Label of image/dog.png is: %d" % lab[0][0]
+
 
 if __name__ == '__main__':
     main()