pip的安装tensorflow【目标检测模块】

╰+攻爆jí腚メ 2021-12-23 03:45 248阅读 0赞

官方教程：[https://github.com/tensorflow/models/blob/master/research/object\_detection/g3doc/installation.md][https_github.com_tensorflow_models_blob_master_research_object_detection_g3doc_installation.md]

官方预训练的模型：[https://github.com/tensorflow/models/blob/master/research/object\_detection/g3doc/detection\_model\_zoo.md][https_github.com_tensorflow_models_blob_master_research_object_detection_g3doc_detection_model_zoo.md]

安装

pip install

卸载

pip uninstall

有的时候升级一个包时，需要先卸载，但是卸载不成功，我们可以直接升级（忽略已经安装的旧版本）

pip install --ignore-installed 包名

安装Tensorflow的GPU版本（[https://tensorflow.google.cn/install/gpu][https_tensorflow.google.cn_install_gpu]，驱动程序的版本和TensorFlow的版本介绍）

pip install tensorflow-gpu

pip install tensorflow-gpu==1.15

安装Tensorflow的CPU版本

pip install tensorflow==1.14

第一步：COCO API的安装

观察在文件夹 models-master/research 文件夹下，没有 COCO API【pycocotools】,

![watermark_type_ZmFuZ3poZW5naGVpdGk_shadow_10_text_aHR0cHM6Ly96aGFuZ3BpbmcuYmxvZy5jc2RuLm5ldA_size_16_color_FFFFFF_t_70][]

下载cocoapi地址：[https://github.com/cocodataset/cocoapi][https_github.com_cocodataset_cocoapi]

![watermark_type_ZmFuZ3poZW5naGVpdGk_shadow_10_text_aHR0cHM6Ly96aGFuZ3BpbmcuYmxvZy5jc2RuLm5ldA_size_16_color_FFFFFF_t_70 1][]

cd cocoapi-master/pythonAPI

在文件夹下cocoapi-master/pythonAPI的内容如下：

![watermark_type_ZmFuZ3poZW5naGVpdGk_shadow_10_text_aHR0cHM6Ly96aGFuZ3BpbmcuYmxvZy5jc2RuLm5ldA_size_16_color_FFFFFF_t_70 2][]

在该文件夹下，执行命令：

make

在没有执行'make'命令之前，在文件夹pycocotools下的内容如下：

![20190910113737757.PNG][]

执行后的命令，在文件夹pycocotools的内容如下，多了一个.c文件和.so文件。

![watermark_type_ZmFuZ3poZW5naGVpdGk_shadow_10_text_aHR0cHM6Ly96aGFuZ3BpbmcuYmxvZy5jc2RuLm5ldA_size_16_color_FFFFFF_t_70 3][]

编译好之后，将文件夹'pycocotools'复制到model-master/research下

cp -r pycocotools <path_to_tensorflow>/models-master/research

执行上述命令后，在models-master/research文件夹下，多了一个pycocotools文件夹，如下图所示：

![watermark_type_ZmFuZ3poZW5naGVpdGk_shadow_10_text_aHR0cHM6Ly96aGFuZ3BpbmcuYmxvZy5jc2RuLm5ldA_size_16_color_FFFFFF_t_70 4][]

第二步：下载安装protocolbuffer

下载地址：[https://github.com/protocolbuffers/protobuf/releases][https_github.com_protocolbuffers_protobuf_releases]

比如，下载3.8.0版本

![watermark_type_ZmFuZ3poZW5naGVpdGk_shadow_10_text_aHR0cHM6Ly96aGFuZ3BpbmcuYmxvZy5jc2RuLm5ldA_size_16_color_FFFFFF_t_70 5][]

点击进入，根据自己的系统选择对应的版本

![watermark_type_ZmFuZ3poZW5naGVpdGk_shadow_10_text_aHR0cHM6Ly96aGFuZ3BpbmcuYmxvZy5jc2RuLm5ldA_size_16_color_FFFFFF_t_70 6][]

下载后，解压

unzip x86_64.zip

解压后，有命令bin/protoc

在没编译之前，在文件夹**object\_detection/protos**下的内容如下：

![20190627152858508.png][]

利用protoc命令编译如下，在编译的时候，要确保当前目录为 /models\_master/research下，执行以下命令：

./bin/protoc object_detection/protos/*.proto --python_out=.

运行结束后，在文件夹 /models\_master/research/object\_detection/protos下的内容如下，多了对应的.py文件：

![watermark_type_ZmFuZ3poZW5naGVpdGk_shadow_10_text_aHR0cHM6Ly96aGFuZ3BpbmcuYmxvZy5jc2RuLm5ldA_size_16_color_FFFFFF_t_70 7][]

在/etc/profile文件中配置环境变量，注意环境变量PYTHONPATH,大致内容如下：

export LD_LIBRARY_PATH=/usr/local/cuda/lib64:/usr/local/opencv/lib:/usr/local/nccl_2.3.7-1+cuda10.0_x8/lib:$LD_LIBRARY_PATH
    export PKG_CONFIG_PATH=/usr/local/opencv/lib/pkgconfig:$PKG_CONFIG_PATH
    JAVA_HOME=/usr/local/java/jdk1.8.0_191
    export JRE_HOME=/usr/local/java/jdk1.8.0_191/jre
    export CLASSPATH=$JAVA_HOME/lib:$JRE_HOME/lib:$CLASSPATH
    export PATH=/usr/local/cuda/bin:$JAVA_HOME/bin:$JRE_HOME/bin:$PATH
    export PYTHONPATH=/home/manager/software/models-master/research:/home/manager/software/models-master/research/slim:$PYTHONPATH

执行以下命令，使其生效:

source /etc/profile

接着执行以下命令:

python object_detection/builders/model_builder_test.py

运行结果如下:

![watermark_type_ZmFuZ3poZW5naGVpdGk_shadow_10_text_aHR0cHM6Ly96aGFuZ3BpbmcuYmxvZy5jc2RuLm5ldA_size_16_color_FFFFFF_t_70 8][]

代表运行成功

[https_github.com_tensorflow_models_blob_master_research_object_detection_g3doc_installation.md]: https://github.com/tensorflow/models/blob/master/research/object_detection/g3doc/installation.md
[https_github.com_tensorflow_models_blob_master_research_object_detection_g3doc_detection_model_zoo.md]: https://github.com/tensorflow/models/blob/master/research/object_detection/g3doc/detection_model_zoo.md
[https_tensorflow.google.cn_install_gpu]: https://tensorflow.google.cn/install/gpu
[watermark_type_ZmFuZ3poZW5naGVpdGk_shadow_10_text_aHR0cHM6Ly96aGFuZ3BpbmcuYmxvZy5jc2RuLm5ldA_size_16_color_FFFFFF_t_70]: /images/20211223/033c9ffbfe81401d80187636f477036b.png
[https_github.com_cocodataset_cocoapi]: https://github.com/cocodataset/cocoapi
[watermark_type_ZmFuZ3poZW5naGVpdGk_shadow_10_text_aHR0cHM6Ly96aGFuZ3BpbmcuYmxvZy5jc2RuLm5ldA_size_16_color_FFFFFF_t_70 1]: /images/20211223/149c34adb4e64380a468a3d5e8a93609.png
[watermark_type_ZmFuZ3poZW5naGVpdGk_shadow_10_text_aHR0cHM6Ly96aGFuZ3BpbmcuYmxvZy5jc2RuLm5ldA_size_16_color_FFFFFF_t_70 2]: /images/20211223/7a70324a6bfc4f14a42c17ccbd368f7f.png
[20190910113737757.PNG]: /images/20211223/a36eee181be548d6b1208038a514d493.png
[watermark_type_ZmFuZ3poZW5naGVpdGk_shadow_10_text_aHR0cHM6Ly96aGFuZ3BpbmcuYmxvZy5jc2RuLm5ldA_size_16_color_FFFFFF_t_70 3]: /images/20211223/df671f49032b48709045040398427f17.png
[watermark_type_ZmFuZ3poZW5naGVpdGk_shadow_10_text_aHR0cHM6Ly96aGFuZ3BpbmcuYmxvZy5jc2RuLm5ldA_size_16_color_FFFFFF_t_70 4]: /images/20211223/4eea94d5a7ee418f8aeff07a7bf0dc31.png
[https_github.com_protocolbuffers_protobuf_releases]: https://github.com/protocolbuffers/protobuf/releases
[watermark_type_ZmFuZ3poZW5naGVpdGk_shadow_10_text_aHR0cHM6Ly96aGFuZ3BpbmcuYmxvZy5jc2RuLm5ldA_size_16_color_FFFFFF_t_70 5]: /images/20211223/c23c03e2c52941b1a83b7f37a4b6b910.png
[watermark_type_ZmFuZ3poZW5naGVpdGk_shadow_10_text_aHR0cHM6Ly96aGFuZ3BpbmcuYmxvZy5jc2RuLm5ldA_size_16_color_FFFFFF_t_70 6]: /images/20211223/52a48d78b6ea437ca72db1e369ce5d67.png
[20190627152858508.png]: /images/20211223/9b661a8681d64cdc9761fe6c72b398d1.png
[watermark_type_ZmFuZ3poZW5naGVpdGk_shadow_10_text_aHR0cHM6Ly96aGFuZ3BpbmcuYmxvZy5jc2RuLm5ldA_size_16_color_FFFFFF_t_70 7]: /images/20211223/819e18c240e24011aa3b336b276fecc7.png
[watermark_type_ZmFuZ3poZW5naGVpdGk_shadow_10_text_aHR0cHM6Ly96aGFuZ3BpbmcuYmxvZy5jc2RuLm5ldA_size_16_color_FFFFFF_t_70 8]: /images/20211223/f1f71908bfc54a3e9575ce1bd4108692.png