运行Tensorflow Object Detection

参考： http://gitlab.hutaojie.com/qiezi/models/blob/master/research/object_detection/g3doc/installation.md

环境：

python 2.7.12

tensorflow 1.8.0

ubuntu 16.04

1. 下载models

git clone https://github.com/tensorflow/models.git

这里直接放在/iris目录下　　

2. 安装依赖库

Tensorflow Object Detection API依赖以下库:

Protobuf 2.6
Python-tk
Pillow 1.0
lxml
tf Object_detection(which is included in the "/iris/models/research/" checkout)
Jupyter notebook
Matplotlib
Tensorflow
Cython
cocoapi

3. 如何安装Tensorflow , 请参考：

windows：https://www.cnblogs.com/lixiaolun/p/9031997.html

ubuntu：https://www.cnblogs.com/lixiaolun/p/9024822.html

我这里直接pip安装（已经装好python）：

# For CPU
pip install tensorflow
# For GPU
pip install tensorflow-gpu

4. 其他库使用一下命令安装：

sudo apt-get install protobuf-compiler python-pil python-lxml python-tk
sudo pip install Cython
sudo pip install jupyter
sudo pip install matplotlib

可以使用pip安装依赖的库:

sudo pip install Cython
sudo pip install pillow
sudo pip install lxml
sudo pip install jupyter
sudo pip install matplotlib

5. 安装COCO API

下载并编译COCO API

git clone https://github.com/cocodataset/cocoapi.git
cd cocoapi/PythonAPI
make

然后拷贝pycocotools 目录到/root/models/research目录下

cp -r pycocotools /iris/models/research/

6. Protobuf 编译

cd /iris/models/research/
protoc object_detection/protos/*.proto --python_out=.

注： protoc需要3.*版本，如下安装请参考：https://github.com/tensorflow/models/issues/4002

# 1. Make sure you grab the latest version
curl -OL https://github.com/google/protobuf/releases/download/v3.2.0/protoc-3.2.0-linux-x86_64.zip

# 2. Unzip
unzip protoc-3.2.0-linux-x86_64.zip -d protoc3

# 3. change your protobuf to protobuf3

# 3.1 Move protoc to /usr/local/bin/
sudo mv protoc3/bin/* /usr/local/bin/

# 3.2 Move protoc3/include to /usr/local/include/
sudo mv protoc3/include/* /usr/local/include/

# 4. validate protoc version
protoc --version

7. 添加库到PYTHONPATH

cd /iris/models/research/
export PYTHONPATH=$PYTHONPATH:`pwd`:`pwd`/slim:`pwd`/object_detection

注：这条命令需要在每一个终端都执行以下，当然也可以加入到~/.bashrc文件中。

8. 测试安装结果

python object_detection/builders/model_builder_test.py

运行Tensorflow Object Detection

猜你喜欢