使用的系统：ubuntu 16.04 LTS

安装

1. 安装conda, 需要输入的地方输入yes

curl -O https://repo.continuum.io/archive/Anaconda3-5.0.1-Linux-x86_64.sh
bash Anaconda3-5.0.1-Linux-x86_64.sh

创建conda环境, 记得新开窗口

conda create -n tensorWork pip python=3.5

进入conda环境

source activate tensorWork

2. 安装tensorflow pip install tensorflow

3. 安装其他组件

conda install -c anaconda protobuf
pip install pillow lxml Cython contextlib2 jupyter matplotlib pandas opencv-python ConfigParser nets  

4. 安装物体识别

git clone https://github.com/tensorflow/models.git
cd models/research
protoc --python_out=. ./object_detection/protos/anchor_generator.proto ./object_detection/protos/argmax_matcher.proto ./object_detection/protos/bipartite_matcher.proto ./object_detection/protos/box_coder.proto ./object_detection/protos/box_predictor.proto ./object_detection/protos/eval.proto ./object_detection/protos/faster_rcnn.proto ./object_detection/protos/faster_rcnn_box_coder.proto ./object_detection/protos/grid_anchor_generator.proto ./object_detection/protos/hyperparams.proto ./object_detection/protos/image_resizer.proto ./object_detection/protos/input_reader.proto ./object_detection/protos/losses.proto ./object_detection/protos/matcher.proto ./object_detection/protos/mean_stddev_box_coder.proto ./object_detection/protos/model.proto ./object_detection/protos/optimizer.proto ./object_detection/protos/pipeline.proto ./object_detection/protos/post_processing.proto ./object_detection/protos/preprocessor.proto ./object_detection/protos/region_similarity_calculator.proto ./object_detection/protos/square_box_coder.proto ./object_detection/protos/ssd.proto ./object_detection/protos/ssd_anchor_generator.proto ./object_detection/protos/string_int_label_map.proto ./object_detection/protos/train.proto ./object_detection/protos/keypoint_box_coder.proto ./object_detection/protos/multiscale_anchor_generator.proto ./object_detection/protos/graph_rewriter.proto ./object_detection/protos/calibration.proto ./object_detection/protos/flexible_grid_anchor_generator.proto

python setup.py build

python setup.py install

vim ~/.bashrc
添加环境变量。相关路径修改成自己的
export PYTHONPATH="/home/wayne/Work/models:/home/wayne/Work/models/research:/home/wayne/Work/models/research/slim:$PYTHONPATH"
打开新窗口，重新接入
source activate tensorWork

5. 安装标注工具

git clone https://github.com/tzutalin/labelImg.git
sudo apt-get install pyqt5-dev-tools
pip install pyqt5==5.10.1 lxml==4.2.4
cd labelImg/
make qt5py3

训练

1. 进入环境

source activate tensorWork

2. 创建训练项目example

mkdir tensorWork/datas/example

3. 图片预处理 进行图片压缩

python tensorWork/resizer.py path/to/image/dir

第二个参数为图片所在文件夹

4. 添加图片

mkdir tensorWork/datas/example/images/train
mkdir tensorWork/datas/example/images/test
mkdir tensorWork/datas/example/images/valid

train 下放需要训练的图片 80%， 需要标注 test 下放需要训练测试的图片 20%， 需要标注 valid 下放用于验证训练结果的图片，不同于train test的图片

5. 图片标注 python labelImg/labelImg.py

选择对应文件夹 train test 进行标注

6. 添加配置文件 tensorWork/datas/example/config.conf

配置内容

[base]
num_classes=2
num_steps=1000
[label]
1=pan
2=luoshuan

num_classes 表示要训练的分类数量， 与 label中的数量相等 num_steps 表示训练的步数 label中为图片标注时用的label名称，按序填写

7. 进行训练 python tensorWork/main.py tensorWork/datas/example

8. 如果训练完成1000次后， 想继续训练至2000次，只需修改num_steps=2000，再启动训练

识别

1. 需要识别的图片放到tensorWork/datas/images/valid中
2. python tensorWork/valid.py tensorWork/datas/example
3. esc退出