diff --git a/example/mobilenetv2/Readme.md b/example/mobilenetv2/Readme.md
index 4a1b8c26a1..9427b1a6b7 100644
--- a/example/mobilenetv2/Readme.md
+++ b/example/mobilenetv2/Readme.md
@@ -8,11 +8,11 @@ MobileNetV2 builds upon the ideas from MobileNetV1, using depthwise separable co
 
 # Dataset
 
-Dataset used: imagenet
+Dataset used: imagenet2012
 
-- Dataset size: ~125G, 1.2W colorful images in 1000 classes
-	- Train: 120G, 1.2W images
-	- Test: 5G, 50000 images
+- Dataset size: ~125G
+	- Train: 120G, 1281167 images: 1000 directories
+	- Test: 5G, 50000 images: images should be classified into 1000 directories firstly, just like train images
 - Data format: RGB images.
 	- Note: Data will be processed in src/dataset.py 
 
@@ -139,4 +139,4 @@ result: {'acc': 0.71976314102564111} ckpt=/path/to/checkpoint/mobilenet-200_625.
 | Model for inference        |                               |                           |                      |
 
 # ModelZoo Homepage  
- [Link](https://gitee.com/mindspore/mindspore/tree/master/mindspore/model_zoo)  
\ No newline at end of file
+ [Link](https://gitee.com/mindspore/mindspore/tree/master/mindspore/model_zoo)
diff --git a/example/mobilenetv2_quant/Readme.md b/example/mobilenetv2_quant/Readme.md
index 361b74dd12..f426302454 100644
--- a/example/mobilenetv2_quant/Readme.md
+++ b/example/mobilenetv2_quant/Readme.md
@@ -10,9 +10,9 @@ MobileNetV2 builds upon the ideas from MobileNetV1, using depthwise separable co
 
 Dataset used: imagenet
 
-- Dataset size: ~125G, 1.2W colorful images in 1000 classes
-	- Train: 120G, 1.2W images
-	- Test: 5G, 50000 images
+- Dataset size: ~125G
+	- Train: 120G, 1281167 images: 1000 directories
+	- Test: 5G, 50000 images: images should be classified into 1000 directories firstly, just like train images
 - Data format: RGB images.
 	- Note: Data will be processed in src/dataset.py 
 
@@ -99,4 +99,4 @@ result: {'acc': 0.71976314102564111} ckpt=/path/to/checkpoint/mobilenet-200_625.
 
 
 # ModelZoo Homepage  
- [Link](https://gitee.com/mindspore/mindspore/tree/master/mindspore/model_zoo)  
\ No newline at end of file
+ [Link](https://gitee.com/mindspore/mindspore/tree/master/mindspore/model_zoo)
diff --git a/example/resnet50_imagenet2012/README.md b/example/resnet50_imagenet2012/README.md
index 6baf863544..a5913c233d 100644
--- a/example/resnet50_imagenet2012/README.md
+++ b/example/resnet50_imagenet2012/README.md
@@ -14,7 +14,7 @@ This is an example of training ResNet-50 with ImageNet2012 dataset in MindSpore.
 > ```
 > .  
 > ├── ilsvrc                  # train dataset
-> └── ilsvrc_eval             # infer dataset
+> └── ilsvrc_eval             # infer dataset: images should be classified into 1000 directories firstly, just like train images
 > ```
 
 
@@ -147,4 +147,4 @@ python train.py --dataset_path=dataset/ilsvrc/train --device_target="GPU" --pre_
 
 # infer example
 python eval.py --dataset_path=dataset/ilsvrc/val --device_target="GPU" --checkpoint_path=resnet-90_5004ss.ckpt
-```
\ No newline at end of file
+```
diff --git a/example/resnet50_imagenet2012_THOR/README.md b/example/resnet50_imagenet2012_THOR/README.md
index 6003d8d7b7..b1ad840d57 100644
--- a/example/resnet50_imagenet2012_THOR/README.md
+++ b/example/resnet50_imagenet2012_THOR/README.md
@@ -14,7 +14,7 @@ This is an example of training ResNet-50 V1.5 with ImageNet2012 dataset by secon
 > ```
 > .  
 > ├── ilsvrc                  # train dataset
-> └── ilsvrc_eval             # infer dataset
+> └── ilsvrc_eval             # infer dataset: images should be classified into 1000 directories firstly, just like train images
 > ```
 
 
diff --git a/example/resnet50_quant/README.md b/example/resnet50_quant/README.md
index f15ef2d59d..948bd93ec6 100644
--- a/example/resnet50_quant/README.md
+++ b/example/resnet50_quant/README.md
@@ -14,7 +14,7 @@ This is an example of training ResNet-50_quant with ImageNet2012 dataset in Mind
 > ```
 > .  
 > ├── ilsvrc                  # train dataset
-> └── ilsvrc_eval             # infer dataset
+> └── ilsvrc_eval             # infer dataset: images should be classified into 1000 directories firstly, just like train images
 > ```