Tensorflow Object Detection API
程序员文章站
2024-03-14 10:05:10
...
1 首先下载源码
https://github.com/tensorflow/models
2 按照官方说明文档安装 依赖库:
Tensorflow Object Detection API depends on the following libraries:
Protobuf 2.6
Pillow 1.0
lxml
tf Slim (which is included in the “tensorflow/models” checkout)
Jupyter notebook
Matplotlib
Tensorflow
(一般地:直接安装anacnonda基本这些库都会有,不过要手动安装tensoflow)
3 配置环境变量
models/research/ 和 slim 目录需要添加进 PYTHONPATH:
d:\tensorflow\models\research
d:\tensorflow\models\research\slim
(斜体部分按照你放置models文件夹的盘符来定。)
4 安装protoc
我下载的版本是protoc-3.3.0-win32.zip,解压后将bin文件夹中的【protoc.exe】放到C:\Windows
5 编译proto模型(重点)
进入目录:**/models/research/
在命令行下执行下面命令,protobuf 2.6不再支持文件名通配符,吐一下血,建议做个.bat文件,批量执行下述命令,同时将–proto_path中指定的目录添加进PATH环境变量。:
【D:\tensorflow\要换成你的电脑放置models文件夹的盘符位置】
protoc --proto_path=D:\tensorflow\models\research\ --python_out=. D:\tensorflow\models\research\object_detection\protos\anchor_generator.proto
protoc --proto_path=D:\tensorflow\models\research\ --python_out=. D:\tensorflow\models\research\object_detection\protos\argmax_matcher.proto
protoc --proto_path=D:\tensorflow\models\research\ --python_out=. D:\tensorflow\models\research\object_detection\protos\bipartite_matcher.proto
protoc --proto_path=D:\tensorflow\models\research\ --python_out=. D:\tensorflow\models\research\object_detection\protos\box_coder.proto
protoc --proto_path=D:\tensorflow\models\research\ --python_out=. D:\tensorflow\models\research\object_detection\protos\box_predictor.proto
protoc --proto_path=D:\tensorflow\models\research\ --python_out=. D:\tensorflow\models\research\object_detection\protos\eval.proto
protoc --proto_path=D:\tensorflow\models\research\ --python_out=. D:\tensorflow\models\research\object_detection\protos\faster_rcnn.proto
protoc --proto_path=D:\tensorflow\models\research\ --python_out=. D:\tensorflow\models\research\object_detection\protos\faster_rcnn_box_coder.proto
protoc --proto_path=D:\tensorflow\models\research\ --python_out=. D:\tensorflow\models\research\object_detection\protos\grid_anchor_generator.proto
protoc --proto_path=D:\tensorflow\models\research\ --python_out=. D:\tensorflow\models\research\object_detection\protos\hyperparams.proto
protoc --proto_path=D:\tensorflow\models\research\ --python_out=. D:\tensorflow\models\research\object_detection\protos\image_resizer.proto
protoc --proto_path=D:\tensorflow\models\research\ --python_out=. D:\tensorflow\models\research\object_detection\protos\input_reader.proto
protoc --proto_path=D:\tensorflow\models\research\ --python_out=. D:\tensorflow\models\research\object_detection\protos\keypoint_box_coder.proto
protoc --proto_path=D:\tensorflow\models\research\ --python_out=. D:\tensorflow\models\research\object_detection\protos\losses.proto
protoc --proto_path=D:\tensorflow\models\research\ --python_out=. D:\tensorflow\models\research\object_detection\protos\matcher.proto
protoc --proto_path=D:\tensorflow\models\research\ --python_out=. D:\tensorflow\models\research\object_detection\protos\mean_stddev_box_coder.proto
protoc --proto_path=D:\tensorflow\models\research\ --python_out=. D:\tensorflow\models\research\object_detection\protos\model.proto
protoc --proto_path=D:\tensorflow\models\research\ --python_out=. D:\tensorflow\models\research\object_detection\protos\optimizer.proto
protoc --proto_path=D:\tensorflow\models\research\ --python_out=. D:\tensorflow\models\research\object_detection\protos\pipeline.proto
protoc --proto_path=D:\tensorflow\models\research\ --python_out=. D:\tensorflow\models\research\object_detection\protos\post_processing.proto
protoc --proto_path=D:\tensorflow\models\research\ --python_out=. D:\tensorflow\models\research\object_detection\protos\preprocessor.proto
protoc --proto_path=D:\tensorflow\models\research\ --python_out=. D:\tensorflow\models\research\object_detection\protos\region_similarity_calculator.proto
protoc --proto_path=D:\tensorflow\models\research\ --python_out=. D:\tensorflow\models\research\object_detection\protos\square_box_coder.proto
protoc --proto_path=D:\tensorflow\models\research\ --python_out=. D:\tensorflow\models\research\object_detection\protos\ssd.proto
protoc --proto_path=D:\tensorflow\models\research\ --python_out=. D:\tensorflow\models\research\object_detection\protos\ssd_anchor_generator.proto
protoc --proto_path=D:\tensorflow\models\research\ --python_out=. D:\tensorflow\models\research\object_detection\protos\string_int_label_map.proto
protoc --proto_path=D:\tensorflow\models\research\ --python_out=. D:\tensorflow\models\research\object_detection\protos\train.proto
6 检测模型安装成功
在 object_detection/builders/目录下,cmd命令行运行
运行python model_builder_test.py,检测是否安装成功
推荐阅读
-
Tensorflow Object Detection API
-
Tensorflow Object Detection API
-
TensorFlow Object Detection API
-
ubuntu下谷歌开源的TensorFlow Object Detection API的安装教程
-
tensorflow入门教程(二十五)Object Detection API目标检测(下)
-
[论文复现] Faster R-CNN: Towards Real-Time Object Detection with Region Proposal Networks
-
ImageAI (二) 使用Python快速简单实现物体检测 Object Detection
-
tensorflow入门教程(二十四)Object Detection API目标检测(中)
-
windows10 tensorflow_(object_detection) 实现三(快速上手)
-
使用tensorflow object_detection API完成目标检测