Faster rcnn排坑指南

亦凉 2021-11-01 15:26 275阅读 0赞

1、在全球最大搅基论坛下载Faster rcnn代码（代码基于3.5，但我的环境为py3.6）：[https://github.com/dBeker/Faster-RCNN-TensorFlow-Python3.5][https_github.com_dBeker_Faster-RCNN-TensorFlow-Python3.5]

2、修改`Faster-RCNN-TensorFlow-Python3.5-master\data/coco/PythonAPI/setup.py`文件：在第15行加上

,
        Extension( 'lib.utils.cython_bbox',
                   sources=['../../../lib/utils/bbox.c','../../../lib/utils/bbox.pyx'],
                   include_dirs = [np.get_include(), '/lib/utils'], 
                   extra_compile_args=[], )

由于没有bbox.c和blob.py文件。所以要先在Faster-RCNN-TensorFlow-Python3.5-master\\lib\\utils执行

python setup.py build\_ext --inplace  
生成cython\_bbox.c和cython\_bbox.pyx，然后将这两个改名为bbox.c和bbox.pyx。

再在./data/coco/PythonAPI下面运行

python setup.py build_ext --inplace
    
     python setup.py build_ext install

下载数据集

http://host.robots.ox.ac.uk/pascal/VOC/voc2007/VOCtrainval_06-Nov-2007.tar
    http://host.robots.ox.ac.uk/pascal/VOC/voc2007/VOCtest_06-Nov-2007.tar
    http://host.robots.ox.ac.uk/pascal/VOC/voc2007/VOCdevkit_08-Jun-2007.tar

将三个文件同时解压到VOCdevkit2007目录下，并将VOCdevkit2007目录存放于data目录下

下载VGG模型：[http://download.tensorflow.org/models/vgg\_16\_2016\_08\_28.tar.gz][http_download.tensorflow.org_models_vgg_16_2016_08_28.tar.gz]

解压并改名为vgg16.ckpt(原来是vgg\_16.ckpt)后存放在`Faster-RCNN-TensorFlow-Python3.5-master\data\imagenet_weights\`目录下

都弄好以后，就可以训练了执行:

python train.py

转载于:https://www.cnblogs.com/roscangjie/p/10945001.html

[https_github.com_dBeker_Faster-RCNN-TensorFlow-Python3.5]: https://github.com/dBeker/Faster-RCNN-TensorFlow-Python3.5
[http_download.tensorflow.org_models_vgg_16_2016_08_28.tar.gz]: http://download.tensorflow.org/models/vgg_16_2016_08_28.tar.gz