Tensorflow Object Detection Api自定义训练

# 创建并且激活环境
conda create -n vott python=3.7
conda activate vott
# 临时更换清华pip镜像源
pip install -i https://pypi.tuna.tsinghua.edu.cn/simple some-package
pip install --upgrade tensorflow-gpu==1.14
# 测试TensorFlow安装
python
>>>import tensorflow as tf
>>>tf.__version__
>>>tf.__path__
>>>hello=tf.constant(“hello tensorflow”)
>>>sess=tf.Session()
>>>print(sess.run(hello))
# 安装依赖包
conda install pillow=6.2.1 lxml=4.4.1 jupyter=1.0.0 matplotlib=3.1.1 opencv=3.4.2 pathlib=1.0.1 cython

下载安装TensorFlow模型

下载链接，选择1.14.0版本的 Tensorflow Models releases，并且解压

Protobuf安装编译

下载链接，选择3.12.3版本的 protoc releases page

# 添加protobuf环境变量
vim  ~/.bashrc

在~/.bashrc文件中添加以下内容，这里换成自己的protobuf路径

export PATH=/home/dapeng/PycharmProjects/TensorFlow/protobuf-3.12.3/:$PATH

# 更新protobuf环境变量
source  ~/.bashrc
# <PATH_TO_TF>指的是/research/的绝对路径
cd <PATH_TO_TF>/TensorFlow/models/research/
# From within TensorFlow/models/research/
protoc object_detection/protos/*.proto --python_out=.
# From within TensorFlow/models/research/
pip install .
# 添加research/slim到PYTHONPATH环境变量
vim  ~/.bashrc

在~/.bashrc文件中添加以下内容

# From within tensorflow/models/research/
export PYTHONPATH=$PYTHONPATH:<PATH_TO_TF>/TensorFlow/models/research/slim
# example: export PYTHONPATH=~/project/caffe/python:/home/dapeng/PycharmProjects/TensorFlow/models/research/:/home/dapeng/PycharmProjects/TensorFlow/models/research/slim:$PYTHONPATH

COCO API安装编译

git clone https://github.com/cocodataset/cocoapi.git
cd cocoapi/PythonAPI
make
# <PATH_TO_TF>指的是/research/的绝对路径
cp -r pycocotools <PATH_TO_TF>/TensorFlow/models/research/

测试安装

cd <PATH_TO_TF>/TensorFlow/models/research/object_detection
# From within TensorFlow/models/research/object_detection
jupyter notebook

点击‘运行所有’，等10分钟，红色区域不算报错，没关系的

大功告成！！！！！！！！！！！！