未验证 提交 d8edeb3b 编写于 作者: H huangjianhui 提交者: GitHub

Merge branch 'develop' into develop

# PP-Shitu # C++ PP-Shitu Service
**注意:** Since opencv library is used in the pre-processing and post-processing of C++ code, you need to open `WITH_OPENCV` option to compile the C++ server, see [COMPILE.md](../../../../doc/Compile_CN.md)
```
# Dependent packages need to be installed
pip install faiss-cpu==1.7.1.post2 -i https://mirror.baidu.com/pypi/simple
```
## Get Model ## Get Model
``` ```
wget https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/rec/models/inference/serving/pp_shitu.tar.gz wget https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/rec/models/inference/serving/pp_shitu.tar.gz
tar -xzvf pp_shitu.tar.gz tar -xvf pp_shitu.tar.gz
``` ```
## Get test images and index ## Get test images and index
...@@ -20,5 +27,5 @@ sh run_cpp_serving.sh ...@@ -20,5 +27,5 @@ sh run_cpp_serving.sh
### Client Prediction ### Client Prediction
``` ```
python3 test_cpp_serving_pipeline.py ./drint_dataset_v1.0/test_images/nongfu_spring.jpeg python3 test_cpp_serving_pipeline.py ./drink_dataset_v1.0/test_images/nongfu_spring.jpeg
``` ```
# PP-Shitu # C++ PP-Shitu Service服务
**注意:** 由于C++代码中前后处理中使用了OPENCV库,您需要使用开启WITH_OPENCV选项编译的C++ Server,详见[COMPILE.md](../../../../doc/Compile_CN.md)
```
## 需安装相关依赖包
pip install faiss-cpu==1.7.1.post2 -i https://mirror.baidu.com/pypi/simple
```
## 获取模型 ## 获取模型
``` ```
wget https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/rec/models/inference/serving/pp_shitu.tar.gz wget https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/rec/models/inference/serving/pp_shitu.tar.gz
tar -xzvf pp_shitu.tar.gz tar -xvf pp_shitu.tar.gz
``` ```
## 获取测试图像和index ## 获取测试图像和index
...@@ -20,5 +27,5 @@ sh run_cpp_serving.sh ...@@ -20,5 +27,5 @@ sh run_cpp_serving.sh
### 客户端预测 ### 客户端预测
``` ```
python3 test_cpp_serving_pipeline.py ./drint_dataset_v1.0/test_images/nongfu_spring.jpeg python3 test_cpp_serving_pipeline.py ./drink_dataset_v1.0/test_images/nongfu_spring.jpeg
``` ```
Markdown is supported
0% .
You are about to add 0 people to the discussion. Proceed with caution.
先完成此消息的编辑!
想要评论请 注册