diff --git a/README_ch.md b/README_ch.md
index 45abc1bdd0570159721d9c14b727a276a192fc85..61335ae0ef00a76dabd978ad31ff5d8cf751b218 100644
--- a/README_ch.md
+++ b/README_ch.md
@@ -7,6 +7,8 @@
 飞桨图像识别套件PaddleClas是飞桨为工业界和学术界所准备的一个图像识别任务的工具集，助力使用者训练出更好的视觉模型和应用落地。
 
 **近期更新**
+- 2021.10.31 优化文档，新增饮料识别demo
+- 2021.10.23 新增轻量级检测、特征提取模型，新增DeepHash模块，检索模块切换为faiss，支持PaddleServing和PaddleSlim
 - 2021.09.17 增加PaddleClas自研PP-LCNet系列模型, 这些模型在Intel CPU上有较强的竞争力。PP-LCNet的介绍可以参考[论文](https://arxiv.org/pdf/2109.15099.pdf)或者[PP-LCNet模型介绍](docs/zh_CN/models/PP-LCNet.md)，相关指标和预训练权重可以从 [这里](docs/zh_CN/ImageNet_models_cn.md)下载。
 - 2021.08.11 更新7个[FAQ](docs/zh_CN/faq_series/faq_2021_s2.md)。
 - 2021.06.29 添加Swin-transformer系列模型，ImageNet1k数据集上Top1 acc最高精度可达87.2%；支持训练预测评估与whl包部署，预训练模型可以从[这里](docs/zh_CN/models/models_intro.md)下载。
@@ -89,7 +91,7 @@ Res2Net200_vd预训练模型Top-1精度高达85.1%。
 ## 图像识别系统介绍
 
 <div align="center">
-<img src="./docs/images/structure.png"  width = "400" />
+<img src="./docs/images/structure.jpg"  width = "800" />
 </div>
 
 整个图像识别系统分为三步：（1）通过一个目标检测模型，检测图像物体候选区域（2）对每个候选区域进行特征提取（3）与检索库中图像进行特征匹配，提取识别结果。
@@ -99,6 +101,11 @@ Res2Net200_vd预训练模型Top-1精度高达85.1%。
 
 <a name="识别效果展示"></a>
 ## 更多效果展示 [more](https://github.com/PaddlePaddle/PaddleClas/tree/release/2.2/docs/images/recognition/more_demo_images)
+- 瓶装饮料识别
+<div align="center">
+<img src="docs/images/drink_demo.gif">
+</div>
+
 - 商品识别
 <div align="center">
 <img src="https://user-images.githubusercontent.com/18028216/122769644-51604f80-d2d7-11eb-8290-c53b12a5c1f6.gif"  width = "400" />
diff --git a/README_en.md b/README_en.md
index aa65ce3825ac9a2d5691be6a1dcc45aae1004626..b6dbd1281822215a8d1cbef8f6dff658dd019f2c 100644
--- a/README_en.md
+++ b/README_en.md
@@ -83,7 +83,7 @@ Quick experience of image recognition：[Link](./docs/en/tutorials/quick_start_r
 ## Introduction to Image Recognition Systems
 
 <div align="center">
-<img src="./docs/images/mainpage/recognition_pipeline_en.png"  width = "400" />
+<img src="./docs/images/structure.jpg"  width = "800" />
 </div>
 
 Image recognition can be divided into three steps:
diff --git a/deploy/configs/build_drinks.yaml b/deploy/configs/build_drinks.yaml
new file mode 100644
index 0000000000000000000000000000000000000000..79735e1a17e4529a0eb0af4b323b5a2c9ae6b695
--- /dev/null
+++ b/deploy/configs/build_drinks.yaml
@@ -0,0 +1,36 @@
+Global:
+  rec_inference_model_dir: "./models/general_PPLCNet_x2_5_lite_v1.0_infer"
+  batch_size: 32
+  use_gpu: True
+  enable_mkldnn: False
+  cpu_num_threads: 10
+  enable_benchmark: False
+  use_fp16: False
+  ir_optim: True
+  use_tensorrt: False
+  gpu_mem: 8000
+  enable_profile: False
+
+RecPreProcess:
+  transform_ops:
+    - ResizeImage:
+        size: 224
+    - NormalizeImage:
+        scale: 0.00392157
+        mean: [0.485, 0.456, 0.406]
+        std: [0.229, 0.224, 0.225]
+        order: ''
+    - ToCHWImage:
+
+RecPostProcess: null
+
+# indexing engine config
+IndexProcess:
+  index_method: "HNSW32" # supported: HNSW32, IVF, Flat
+  index_dir: "./drink_dataset/index"
+  image_root: "./drink_dataset"
+  data_file:  "./drink_dataset/drink_label.txt"
+  index_operation: "new" # suported: "append", "remove", "new"
+  delimiter: " "
+  dist_type: "IP"
+  embedding_size: 512
diff --git a/deploy/configs/inference_drink.yaml b/deploy/configs/inference_drink.yaml
new file mode 100644
index 0000000000000000000000000000000000000000..57c0ecb7b10b4adacefef6fe233180d2d8bc4fff
--- /dev/null
+++ b/deploy/configs/inference_drink.yaml
@@ -0,0 +1,55 @@
+Global:
+  infer_imgs: "./recognition_demo_data_v1.2/test_drinks/hongniu_1.jpg"
+  det_inference_model_dir: "./models/picodet_PPLCNet_x2_5_mainbody_lite_v1.0_infer"
+  rec_inference_model_dir: "./models/general_PPLCNet_x2_5_lite_v1.0_infer"
+  rec_nms_thresold: 0.05
+  
+  batch_size: 1
+  image_shape: [3, 640, 640]
+  threshold: 0.2
+  max_det_results: 5
+  labe_list:
+  - foreground
+
+  # inference engine config
+  use_gpu: True
+  enable_mkldnn: False
+  cpu_num_threads: 10
+  enable_benchmark: True
+  use_fp16: False
+  ir_optim: True
+  use_tensorrt: False
+  gpu_mem: 8000
+  enable_profile: False
+
+DetPreProcess:
+  transform_ops:
+    - DetResize:
+        interp: 2
+        keep_ratio: false
+        target_size: [640, 640]
+    - DetNormalizeImage:
+        is_scale: true
+        mean: [0.485, 0.456, 0.406]
+        std: [0.229, 0.224, 0.225]
+    - DetPermute: {}
+DetPostProcess: {}
+
+RecPreProcess:
+  transform_ops:
+    - ResizeImage:
+        size: 224
+    - NormalizeImage:
+        scale: 0.00392157
+        mean: [0.485, 0.456, 0.406]
+        std: [0.229, 0.224, 0.225]
+        order: ''
+    - ToCHWImage:
+
+RecPostProcess: null
+
+# indexing engine config
+IndexProcess:
+  index_dir: "./drink_dataset/index"
+  return_k: 5
+  score_thres: 0.4
diff --git a/docs/images/drink_demo.gif b/docs/images/drink_demo.gif
new file mode 100644
index 0000000000000000000000000000000000000000..599a525f7abf6889f864df4a898b5b8b7d2af55c
Binary files /dev/null and b/docs/images/drink_demo.gif differ
diff --git a/docs/images/structure.jpg b/docs/images/structure.jpg
new file mode 100644
index 0000000000000000000000000000000000000000..1d7f5a17b5377606a0e69a6fb45fdab1652af5d0
Binary files /dev/null and b/docs/images/structure.jpg differ
diff --git a/docs/images/structure.png b/docs/images/structure.png
deleted file mode 100644
index 30081c53eb5183c71d0358da9e2ba8431a920733..0000000000000000000000000000000000000000
Binary files a/docs/images/structure.png and /dev/null differ