maskrcnn无法生成训练文件_PaddlePaddle经典实例分割模型Mask RCNN

简介
区域卷积神经网络（RCNN）系列模型为两阶段目标检测器。通过对图像生成候选区域，提取特征，判别特征类别并修正候选框位置。 RCNN系列目前包含两个代表模型：Faster RCNN，Mask RCNN
Faster RCNN 整体网络可以分为4个主要内容：

1. 基础卷积层。作为一种卷积神经网络目标检测方法，Faster RCNN首先使用一组基础的卷积网络提取图像的特征图。特征图被后续RPN层和全连接层共享。本示例采用ResNet-50作为基础卷积层。
2. 区域生成网络(RPN)。RPN网络用于生成候选区域(proposals)。该层通过一组固定的尺寸和比例得到一组锚点(anchors), 通过softmax判断锚点属于前景或者背景，再利用区域回归修正锚点从而获得精确的候选区域。
3. RoI Align。该层收集输入的特征图和候选区域，将候选区域映射到特征图中并池化为统一大小的区域特征图，送入全连接层判定目标类别, 该层可选用RoIPool和RoIAlign两种方式，在config.py中设置roi_func。
4. 检测层。利用区域特征图计算候选区域的类别，同时再次通过区域回归获得检测框最终的精确位置。

Mask RCNN 扩展自Faster RCNN，是经典的实例分割模型。
Mask RCNN同样为两阶段框架，第一阶段扫描图像生成候选框；第二阶段根据候选框得到分类结果，边界框，同时在原有Faster RCNN模型基础上添加分割分支，得到掩码结果，实现了掩码和类别预测关系的解藕。
本例采用COCO数据集的子集，即随机从COCO大数据集中选取30张图组成coco30数据集,数据存放在rcnn/coco30。 本例采用coco30，可通过修改--data_dir修改训练集.
Mask RCNN结构图如下：

注：本例默认用GPU运行，也可通过设置--use_gpu=False使用CPU设备。

阅读本项目之前建议阅读mask-rcnn原版论文https://arxiv.org/abs/1703.06870 相关优秀博客https://blog.csdn.net/WZZ18191171661/article/details/79453780
本例采用COCO数据集的子集，即随机从COCO大数据集中选取30张图组成coco30数据集,数据存放在rcnn/coco30。 本例采用coco30，可通过修改--data_dir修改训练集.COCO数据集介绍：https://blog.csdn.net/wc781708249/article/details/79603522
文件夹介绍：
1.models：构造模型主体的函数
2.coco30: 训练数据集
3.output: 保存训练模型
4:imagenet_resnet50_fusebn: 存放预训练模型
所有执行文件均在mask-rcnn文件夹内，在了解mask-rcnn基本原理的基础上，建议阅读顺序：
1.models文件夹：存放模型主体构造以及loss构造相关文件。
2.utility.py、config.py: 训练相关参数设置。
3.readers.py: 训练数据格式构造。
4.train.py：训练主体程序。
5.其他相关辅助函数

#安装CocoAPI
#!git clone https://github.com/cocodataset/cocoapi.git
!cd cocoapi/PythonAPI && make install

# install pycocotools to the Python site-packages
python setup.py build_ext install
running build_ext
skipping 'pycocotools/_mask.c' Cython extension (up-to-date)
building 'pycocotools._mask' extension
creating build
creating build/common
creating build/temp.linux-x86_64-3.5
creating build/temp.linux-x86_64-3.5/pycocotools
gcc -pthread -Wno-unused-result -Wsign-compare -DNDEBUG -g -fwrapv -O3 -Wall -Wstrict-prototypes -m64 -fPIC -fPIC -I/opt/conda/envs/python35-paddle120-env/lib/python3.5/site-packages/numpy/core/include -I../common -I/opt/conda/envs/python35-paddle120-env/include/python3.5m -c ../common/maskApi.c -o build/temp.linux-x86_64-3.5/../common/maskApi.o -Wno-cpp -Wno-unused-function -std=c99../common/maskApi.c: In function ‘rleToBbox’:../common/maskApi.c:141:31: warning: ‘xp’ may be used uninitialized in this function [-Wmaybe-uninitialized]
if(j%2==0) xp=x; else if(xp<x) { ys=0; ye=h-1; } ^
gcc -pthread -Wno-unused-result -Wsign-compare -DNDEBUG -g -fwrapv -O3 -Wall -Wstrict-prototypes -m64 -fPIC -fPIC -I/opt/conda/envs/python35-paddle120-env/lib/python3.5/site-packages/numpy/core/include -I../common -I/opt/conda/envs/python35-paddle120-env/include/python3.5m -c pycocotools/_mask.c -o build/temp.linux-x86_64-3.5/pycocotools/_mask.o -Wno-cpp -Wno-unused-function -std=c99
creating build/lib.linux-x86_64-3.5
creating build/lib.linux-x86_64-3.5/pycocotools
gcc -pthread -shared -L/opt/conda/envs/python35-paddle120-env/lib -Wl,-rpath=/opt/conda/envs/python35-paddle120-env/lib,--no-as-needed build/temp.linux-x86_64-3.5/../common/maskApi.o build/temp.linux-x86_64-3.5/pycocotools/_mask.o -L/opt/conda/envs/python35-paddle120-env/lib -lpython3.5m -o build/lib.linux-x86_64-3.5/pycocotools/_mask.cpython-35m-x86_64-linux-gnu.so
running install
running bdist_egg
running egg_info
writing pycocotools.egg-info/PKG-INFO
writing dependency_links to pycocotools.egg-info/dependency_links.txt
writing top-level names to pycocotools.egg-info/top_level.txt
writing requirements to pycocotools.egg-info/requires.txt
reading manifest file 'pycocotools.egg-info/SOURCES.txt'
writing manifest file 'pycocotools.egg-info/SOURCES.txt'
installing library code to build/bdist.linux-x86_64/egg
running install_lib
running build_py
copying pycocotools/coco.py -> build/lib.linux-x86_64-3.5/pycocotools
copying pycocotools/cocoeval.py -> build/lib.linux-x86_64-3.5/pycocotools
copying pycocotools/mask.py -> build/lib.linux-x86_64-3.5/pycocotools
copying pycocotools/__init__.py -> build/lib.linux-x86_64-3.5/pycocotools
creating build/bdist.linux-x86_64
creating build/bdist.linux-x86_64/egg
creating build/bdist.linux-x86_64/egg/pycocotools
copying build/lib.linux-x86_64-3.5/pycocotools/coco.py -> build/bdist.linux-x86_64/egg/pycocotools
copying build/lib.linux-x86_64-3.5/pycocotools/_mask.cpython-35m-x86_64-linux-gnu.so -> build/bdist.linux-x86_64/egg/pycocotools
copying build/lib.linux-x86_64-3.5/pycocotools/cocoeval.py -> build/bdist.linux-x86_64/egg/pycocotools
copying build/lib.linux-x86_64-3.5/pycocotools/mask.py -> build/bdist.linux-x86_64/egg/pycocotools
copying build/lib.linux-x86_64-3.5/pycocotools/__init__.py -> build/bdist.linux-x86_64/egg/pycocotools
byte-compiling build/bdist.linux-x86_64/egg/pycocotools/coco.py to coco.cpython-35.pyc
byte-compiling build/bdist.linux-x86_64/egg/pycocotools/cocoeval.py to cocoeval.cpython-35.pyc
byte-compiling build/bdist.linux-x86_64/egg/pycocotools/mask.py to mask.cpython-35.pyc
byte-compiling build/bdist.linux-x86_64/egg/pycocotools/__init__.py to __init__.cpython-35.pyc
creating stub loader for pycocotools/_mask.cpython-35m-x86_64-linux-gnu.so
byte-compiling build/bdist.linux-x86_64/egg/pycocotools/_mask.py to _mask.cpython-35.pyc
creating build/bdist.linux-x86_64/egg/EGG-INFO
copying pycocotools.egg-info/PKG-INFO -> build/bdist.linux-x86_64/egg/EGG-INFO
copying pycocotools.egg-info/SOURCES.txt -> build/bdist.linux-x86_64/egg/EGG-INFO
copying pycocotools.egg-info/dependency_links.txt -> build/bdist.linux-x86_64/egg/EGG-INFO
copying pycocotools.egg-info/requires.txt -> build/bdist.linux-x86_64/egg/EGG-INFO
copying pycocotools.egg-info/top_level.txt -> build/bdist.linux-x86_64/egg/EGG-INFO
writing build/bdist.linux-x86_64/egg/EGG-INFO/native_libs.txt
zip_safe flag not set; analyzing archive contents...
pycocotools.__pycache__._mask.cpython-35: module references __file__
creating 'dist/pycocotools-2.0-py3.5-linux-x86_64.egg' and adding 'build/bdist.linux-x86_64/egg' to it
removing 'build/bdist.linux-x86_64/egg' (and everything under it)
Processing pycocotools-2.0-py3.5-linux-x86_64.egg
creating /opt/conda/envs/python35-paddle120-env/lib/python3.5/site-packages/pycocotools-2.0-py3.5-linux-x86_64.egg
Extracting pycocotools-2.0-py3.5-linux-x86_64.egg to /opt/conda/envs/python35-paddle120-env/lib/python3.5/site-packages
Adding pycocotools 2.0 to easy-install.pth file
Installed /opt/conda/envs/python35-paddle120-env/lib/python3.5/site-packages/pycocotools-2.0-py3.5-linux-x86_64.egg
Processing dependencies for pycocotools==2.0
Searching for matplotlib==2.2.3
Best match: matplotlib 2.2.3
Adding matplotlib 2.2.3 to easy-install.pth file
Using /opt/conda/envs/python35-paddle120-env/lib/python3.5/site-packages
Searching for Cython==0.29.10
Best match: Cython 0.29.10
Adding Cython 0.29.10 to easy-install.pth file
Installing cythonize script to /opt/conda/envs/python35-paddle120-env/bin
Installing cython script to /opt/conda/envs/python35-paddle120-env/bin
Installing cygdb script to /opt/conda/envs/python35-paddle120-env/bin
Using /opt/conda/envs/python35-paddle120-env/lib/python3.5/site-packages
Searching for setuptools==40.4.3
Best match: setuptools 40.4.3
Adding setuptools 40.4.3 to easy-install.pth file
Installing easy_install script to /opt/conda/envs/python35-paddle120-env/bin
Using /opt/conda/envs/python35-paddle120-env/lib/python3.5/site-packages
Searching for numpy==1.16.4
Best match: numpy 1.16.4
Adding numpy 1.16.4 to easy-install.pth file
Installing f2py3.5 script to /opt/conda/envs/python35-paddle120-env/bin
Installing f2py3 script to /opt/conda/envs/python35-paddle120-env/bin
Installing f2py script to /opt/conda/envs/python35-paddle120-env/bin
Using /opt/conda/envs/python35-paddle120-env/lib/python3.5/site-packages
Searching for pyparsing==2.4.0
Best match: pyparsing 2.4.0
Adding pyparsing 2.4.0 to easy-install.pth file
Using /opt/conda/envs/python35-paddle120-env/lib/python3.5/site-packages
Searching for python-dateutil==2.8.0
Best match: python-dateutil 2.8.0
Adding python-dateutil 2.8.0 to easy-install.pth file
Using /opt/conda/envs/python35-paddle120-env/lib/python3.5/site-packages
Searching for pytz==2019.1
Best match: pytz 2019.1
Adding pytz 2019.1 to easy-install.pth file
Using /opt/conda/envs/python35-paddle120-env/lib/python3.5/site-packages
Searching for six==1.11.0
Best match: six 1.11.0
Adding six 1.11.0 to easy-install.pth file
Using /opt/conda/envs/python35-paddle120-env/lib/python3.5/site-packages
Searching for kiwisolver==1.1.0
Best match: kiwisolver 1.1.0
Adding kiwisolver 1.1.0 to easy-install.pth file
Using /opt/conda/envs/python35-paddle120-env/lib/python3.5/site-packages
Searching for cycler==0.10.0
Best match: cycler 0.10.0
Adding cycler 0.10.0 to easy-install.pth file
Using /opt/conda/envs/python35-paddle120-env/lib/python3.5/site-packages
Finished processing dependencies for pycocotools==2.0
rm -rf build# install pycocotools to the Python site-packages
python setup.py build_ext install
running build_ext
skipping 'pycocotools/_mask.c' Cython extension (up-to-date)
building 'pycocotools._mask' extension
creating build
creating build/common
creating build/temp.linux-x86_64-3.5
creating build/temp.linux-x86_64-3.5/pycocotools
gcc -pthread -Wno-unused-result -Wsign-compare -DNDEBUG -g -fwrapv -O3 -Wall -Wstrict-prototypes -m64 -fPIC -fPIC -I/opt/conda/envs/python35-paddle120-env/lib/python3.5/site-packages/numpy/core/include -I../common -I/opt/conda/envs/python35-paddle120-env/include/python3.5m -c ../common/maskApi.c -o build/temp.linux-x86_64-3.5/../common/maskApi.o -Wno-cpp -Wno-unused-function -std=c99../common/maskApi.c: In function ‘rleToBbox’:../common/maskApi.c:141:31: warning: ‘xp’ may be used uninitialized in this function [-Wmaybe-uninitialized]
if(j%2==0) xp=x; else if(xp<x) { ys=0; ye=h-1; } ^
gcc -pthread -Wno-unused-result -Wsign-compare -DNDEBUG -g -fwrapv -O3 -Wall -Wstrict-prototypes -m64 -fPIC -fPIC -I/opt/conda/envs/python35-paddle120-env/lib/python3.5/site-packages/numpy/core/include -I../common -I/opt/conda/envs/python35-paddle120-env/include/python3.5m -c pycocotools/_mask.c -o build/temp.linux-x86_64-3.5/pycocotools/_mask.o -Wno-cpp -Wno-unused-function -std=c99
creating build/lib.linux-x86_64-3.5
creating build/lib.linux-x86_64-3.5/pycocotools
gcc -pthread -shared -L/opt/conda/envs/python35-paddle120-env/lib -Wl,-rpath=/opt/conda/envs/python35-paddle120-env/lib,--no-as-needed build/temp.linux-x86_64-3.5/../common/maskApi.o build/temp.linux-x86_64-3.5/pycocotools/_mask.o -L/opt/conda/envs/python35-paddle120-env/lib -lpython3.5m -o build/lib.linux-x86_64-3.5/pycocotools/_mask.cpython-35m-x86_64-linux-gnu.so
running install
running bdist_egg
running egg_info
writing pycocotools.egg-info/PKG-INFO
writing dependency_links to pycocotools.egg-info/dependency_links.txt
writing top-level names to pycocotools.egg-info/top_level.txt
writing requirements to pycocotools.egg-info/requires.txt
reading manifest file 'pycocotools.egg-info/SOURCES.txt'
writing manifest file 'pycocotools.egg-info/SOURCES.txt'
installing library code to build/bdist.linux-x86_64/egg
running install_lib
running build_py
copying pycocotools/coco.py -> build/lib.linux-x86_64-3.5/pycocotools
copying pycocotools/cocoeval.py -> build/lib.linux-x86_64-3.5/pycocotools
copying pycocotools/mask.py -> build/lib.linux-x86_64-3.5/pycocotools
copying pycocotools/__init__.py -> build/lib.linux-x86_64-3.5/pycocotools
creating build/bdist.linux-x86_64
creating build/bdist.linux-x86_64/egg
creating build/bdist.linux-x86_64/egg/pycocotools
copying build/lib.linux-x86_64-3.5/pycocotools/coco.py -> build/bdist.linux-x86_64/egg/pycocotools
copying build/lib.linux-x86_64-3.5/pycocotools/_mask.cpython-35m-x86_64-linux-gnu.so -> build/bdist.linux-x86_64/egg/pycocotools
copying build/lib.linux-x86_64-3.5/pycocotools/cocoeval.py -> build/bdist.linux-x86_64/egg/pycocotools
copying build/lib.linux-x86_64-3.5/pycocotools/mask.py -> build/bdist.linux-x86_64/egg/pycocotools
copying build/lib.linux-x86_64-3.5/pycocotools/__init__.py -> build/bdist.linux-x86_64/egg/pycocotools
byte-compiling build/bdist.linux-x86_64/egg/pycocotools/coco.py to coco.cpython-35.pyc
byte-compiling build/bdist.linux-x86_64/egg/pycocotools/cocoeval.py to cocoeval.cpython-35.pyc
byte-compiling build/bdist.linux-x86_64/egg/pycocotools/mask.py to mask.cpython-35.pyc
byte-compiling build/bdist.linux-x86_64/egg/pycocotools/__init__.py to __init__.cpython-35.pyc
creating stub loader for pycocotools/_mask.cpython-35m-x86_64-linux-gnu.so
byte-compiling build/bdist.linux-x86_64/egg/pycocotools/_mask.py to _mask.cpython-35.pyc
creating build/bdist.linux-x86_64/egg/EGG-INFO
copying pycocotools.egg-info/PKG-INFO -> build/bdist.linux-x86_64/egg/EGG-INFO
copying pycocotools.egg-info/SOURCES.txt -> build/bdist.linux-x86_64/egg/EGG-INFO
copying pycocotools.egg-info/dependency_links.txt -> build/bdist.linux-x86_64/egg/EGG-INFO
copying pycocotools.egg-info/requires.txt -> build/bdist.linux-x86_64/egg/EGG-INFO
copying pycocotools.egg-info/top_level.txt -> build/bdist.linux-x86_64/egg/EGG-INFO
writing build/bdist.linux-x86_64/egg/EGG-INFO/native_libs.txt
zip_safe flag not set; analyzing archive contents...
pycocotools.__pycache__._mask.cpython-35: module references __file__
creating 'dist/pycocotools-2.0-py3.5-linux-x86_64.egg' and adding 'build/bdist.linux-x86_64/egg' to it
removing 'build/bdist.linux-x86_64/egg' (and everything under it)
Processing pycocotools-2.0-py3.5-linux-x86_64.egg
creating /opt/conda/envs/python35-paddle120-env/lib/python3.5/site-packages/pycocotools-2.0-py3.5-linux-x86_64.egg
Extracting pycocotools-2.0-py3.5-linux-x86_64.egg to /opt/conda/envs/python35-paddle120-env/lib/python3.5/site-packages
Adding pycocotools 2.0 to easy-install.pth file
Installed /opt/conda/envs/python35-paddle120-env/lib/python3.5/site-packages/pycocotools-2.0-py3.5-linux-x86_64.egg
Processing dependencies for pycocotools==2.0
Searching for matplotlib==2.2.3
Best match: matplotlib 2.2.3
Adding matplotlib 2.2.3 to easy-install.pth file
Using /opt/conda/envs/python35-paddle120-env/lib/python3.5/site-packages
Searching for Cython==0.29.10
Best match: Cython 0.29.10
Adding Cython 0.29.10 to easy-install.pth file
Installing cythonize script to /opt/conda/envs/python35-paddle120-env/bin
Installing cython script to /opt/conda/envs/python35-paddle120-env/bin
Installing cygdb script to /opt/conda/envs/python35-paddle120-env/bin
Using /opt/conda/envs/python35-paddle120-env/lib/python3.5/site-packages
Searching for setuptools==40.4.3
Best match: setuptools 40.4.3
Adding setuptools 40.4.3 to easy-install.pth file
Installing easy_install script to /opt/conda/envs/python35-paddle120-env/bin
Using /opt/conda/envs/python35-paddle120-env/lib/python3.5/site-packages
Searching for numpy==1.16.4
Best match: numpy 1.16.4
Adding numpy 1.16.4 to easy-install.pth file
Installing f2py3.5 script to /opt/conda/envs/python35-paddle120-env/bin
Installing f2py3 script to /opt/conda/envs/python35-paddle120-env/bin
Installing f2py script to /opt/conda/envs/python35-paddle120-env/bin
Using /opt/conda/envs/python35-paddle120-env/lib/python3.5/site-packages
Searching for pyparsing==2.4.0
Best match: pyparsing 2.4.0
Adding pyparsing 2.4.0 to easy-install.pth file
Using /opt/conda/envs/python35-paddle120-env/lib/python3.5/site-packages
Searching for python-dateutil==2.8.0
Best match: python-dateutil 2.8.0
Adding python-dateutil 2.8.0 to easy-install.pth file
Using /opt/conda/envs/python35-paddle120-env/lib/python3.5/site-packages
Searching for pytz==2019.1
Best match: pytz 2019.1
Adding pytz 2019.1 to easy-install.pth file
Using /opt/conda/envs/python35-paddle120-env/lib/python3.5/site-packages
Searching for six==1.11.0
Best match: six 1.11.0
Adding six 1.11.0 to easy-install.pth file
Using /opt/conda/envs/python35-paddle120-env/lib/python3.5/site-packages
Searching for kiwisolver==1.1.0
Best match: kiwisolver 1.1.0
Adding kiwisolver 1.1.0 to easy-install.pth file
Using /opt/conda/envs/python35-paddle120-env/lib/python3.5/site-packages
Searching for cycler==0.10.0
Best match: cycler 0.10.0
Adding cycler 0.10.0 to easy-install.pth file
Using /opt/conda/envs/python35-paddle120-env/lib/python3.5/site-packages
Finished processing dependencies for pycocotools==2.0
rm -rf build

#模型训练
!cd mask-rcnn && python train.py 
   --model_save_dir=output/ 
   --pretrained_model=imagenet_resnet50_fusebn 
   --data_dir=coco30 
   --max_iter=3000 
   --use_gpu=True 
   --MASK_ON=True

----------- Configuration Arguments -----------
MASK_ON: 1
anchor_sizes: [32, 64, 128, 256, 512]
aspect_ratios: [0.5, 1.0, 2.0]
batch_size_per_im: 512
class_num: 81
data_dir: coco30
dataset: coco2017
draw_threshold: 0.8
enable_ce: False
freeze_model_save_dir: freeze_model
im_per_batch: 1
image_path: dataset/coco/val2017
learning_rate: 0.01
log_window: 20
max_iter: 3000
max_size: 1333
model_save_dir: output/
nms_thresh: 0.5
padding_minibatch: False
parallel: False
pixel_means: [102.9801, 115.9465, 122.7717]
pretrained_model: imagenet_resnet50_fusebn
rpn_nms_thresh: 0.7
rpn_stride: [16.0, 16.0]
scales: [800]
score_thresh: 0.05
snapshot_stride: 10000
use_gpu: 1
use_profile: False
use_pyreader: False
variance: [1.0, 1.0, 1.0, 1.0]
------------------------------------------------
W0828 10:16:38.641909 212 http://device_context.cc:259] Please NOTE: device: 0, CUDA Capability: 70, Driver API Version: 9.2, Runtime API Version: 9.0
W0828 10:16:38.646142 212 http://device_context.cc:267] device: 0, cuDNN Version: 7.3.
Creating: coco2017
loading annotations into memory...
Done (t=0.00s)
creating index...
index created!
_add_gt_annotations took 0.009s
Appending horizontally-flipped training examples...
Loaded dataset: coco2017
60 roidb entries
Filtered 0 roidb entries: 60 -> 60
train on coco2017 with 60 roidbs
2019-08-28 10:16:40.598017, iter: 0, lr: 0.00333, 'loss_cls': 4.417, 'loss_mask': 1.109, 'loss_bbox': 0.128, 'loss': 6.569, 'loss_rpn_bbox': 0.216, 'loss_rpn_cls': 0.699, time: 0.018
2019-08-28 10:16:56.832567, iter: 50, lr: 0.00400, 'loss_cls': 0.431, 'loss_mask': 0.704, 'loss_bbox': 0.051, 'loss': 1.744, 'loss_rpn_bbox': 0.146, 'loss_rpn_cls': 0.425, time: 0.268
2019-08-28 10:17:13.274540, iter: 100, lr: 0.00467, 'loss_cls': 0.445, 'loss_mask': 0.656, 'loss_bbox': 0.092, 'loss': 1.783, 'loss_rpn_bbox': 0.205, 'loss_rpn_cls': 0.414, time: 0.297
2019-08-28 10:17:29.150029, iter: 150, lr: 0.00533, 'loss_cls': 0.51, 'loss_mask': 0.645, 'loss_bbox': 0.146, 'loss': 1.855, 'loss_rpn_bbox': 0.165, 'loss_rpn_cls': 0.346, time: 0.300
2019-08-28 10:17:46.438505, iter: 200, lr: 0.00600, 'loss_cls': 0.366, 'loss_mask': 0.623, 'loss_bbox': 0.115, 'loss': 1.583, 'loss_rpn_bbox': 0.156, 'loss_rpn_cls': 0.323, time: 0.283
2019-08-28 10:18:02.917941, iter: 250, lr: 0.00667, 'loss_cls': 0.495, 'loss_mask': 0.612, 'loss_bbox': 0.138, 'loss': 1.737, 'loss_rpn_bbox': 0.155, 'loss_rpn_cls': 0.382, time: 0.314
2019-08-28 10:18:20.530741, iter: 300, lr: 0.00733, 'loss_cls': 0.44, 'loss_mask': 0.591, 'loss_bbox': 0.137, 'loss': 1.616, 'loss_rpn_bbox': 0.132, 'loss_rpn_cls': 0.301, time: 0.358
2019-08-28 10:18:37.671191, iter: 350, lr: 0.00800, 'loss_cls': 0.44, 'loss_mask': 0.564, 'loss_bbox': 0.156, 'loss': 1.59, 'loss_rpn_bbox': 0.139, 'loss_rpn_cls': 0.247, time: 0.377
2019-08-28 10:18:54.597628, iter: 400, lr: 0.00867, 'loss_cls': 0.568, 'loss_mask': 0.591, 'loss_bbox': 0.244, 'loss': 1.738, 'loss_rpn_bbox': 0.104, 'loss_rpn_cls': 0.266, time: 0.349
2019-08-28 10:19:11.880430, iter: 450, lr: 0.00933, 'loss_cls': 0.538, 'loss_mask': 0.568, 'loss_bbox': 0.163, 'loss': 1.59, 'loss_rpn_bbox': 0.112, 'loss_rpn_cls': 0.27, time: 0.338
2019-08-28 10:19:29.210413, iter: 500, lr: 0.01000, 'loss_cls': 0.416, 'loss_mask': 0.543, 'loss_bbox': 0.17, 'loss': 1.557, 'loss_rpn_bbox': 0.105, 'loss_rpn_cls': 0.235, time: 0.337
2019-08-28 10:19:46.603084, iter: 550, lr: 0.01000, 'loss_cls': 0.57, 'loss_mask': 0.569, 'loss_bbox': 0.202, 'loss': 1.676, 'loss_rpn_bbox': 0.107, 'loss_rpn_cls': 0.246, time: 0.335
2019-08-28 10:20:04.189566, iter: 600, lr: 0.01000, 'loss_cls': 0.608, 'loss_mask': 0.512, 'loss_bbox': 0.213, 'loss': 1.727, 'loss_rpn_bbox': 0.053, 'loss_rpn_cls': 0.232, time: 0.336
2019-08-28 10:20:21.582167, iter: 650, lr: 0.01000, 'loss_cls': 0.571, 'loss_mask': 0.461, 'loss_bbox': 0.311, 'loss': 1.773, 'loss_rpn_bbox': 0.167, 'loss_rpn_cls': 0.192, time: 0.346
2019-08-28 10:20:39.112590, iter: 700, lr: 0.01000, 'loss_cls': 0.547, 'loss_mask': 0.462, 'loss_bbox': 0.225, 'loss': 1.477, 'loss_rpn_bbox': 0.086, 'loss_rpn_cls': 0.157, time: 0.348
2019-08-28 10:20:56.749053, iter: 750, lr: 0.01000, 'loss_cls': 0.535, 'loss_mask': 0.471, 'loss_bbox': 0.226, 'loss': 1.483, 'loss_rpn_bbox': 0.084, 'loss_rpn_cls': 0.155, time: 0.338
2019-08-28 10:21:14.991664, iter: 800, lr: 0.01000, 'loss_cls': 0.488, 'loss_mask': 0.544, 'loss_bbox': 0.204, 'loss': 1.657, 'loss_rpn_bbox': 0.107, 'loss_rpn_cls': 0.203, time: 0.377
2019-08-28 10:21:33.416243, iter: 850, lr: 0.01000, 'loss_cls': 0.57, 'loss_mask': 0.464, 'loss_bbox': 0.31, 'loss': 1.638, 'loss_rpn_bbox': 0.077, 'loss_rpn_cls': 0.138, time: 0.380
2019-08-28 10:21:51.716017, iter: 900, lr: 0.01000, 'loss_cls': 0.725, 'loss_mask': 0.515, 'loss_bbox': 0.3, 'loss': 1.88, 'loss_rpn_bbox': 0.087, 'loss_rpn_cls': 0.175, time: 0.549
2019-08-28 10:22:10.090252, iter: 950, lr: 0.01000, 'loss_cls': 0.632, 'loss_mask': 0.464, 'loss_bbox': 0.261, 'loss': 1.581, 'loss_rpn_bbox': 0.048, 'loss_rpn_cls': 0.121, time: 0.364
2019-08-28 10:22:28.691718, iter: 1000, lr: 0.01000, 'loss_cls': 0.595, 'loss_mask': 0.426, 'loss_bbox': 0.315, 'loss': 1.485, 'loss_rpn_bbox': 0.071, 'loss_rpn_cls': 0.115, time: 0.364
2019-08-28 10:22:46.478322, iter: 1050, lr: 0.01000, 'loss_cls': 0.56, 'loss_mask': 0.395, 'loss_bbox': 0.316, 'loss': 1.45, 'loss_rpn_bbox': 0.069, 'loss_rpn_cls': 0.096, time: 0.344
2019-08-28 10:23:04.306094, iter: 1100, lr: 0.01000, 'loss_cls': 0.547, 'loss_mask': 0.416, 'loss_bbox': 0.314, 'loss': 1.533, 'loss_rpn_bbox': 0.098, 'loss_rpn_cls': 0.13, time: 0.333
2019-08-28 10:23:22.653763, iter: 1150, lr: 0.01000, 'loss_cls': 0.433, 'loss_mask': 0.393, 'loss_bbox': 0.318, 'loss': 1.288, 'loss_rpn_bbox': 0.046, 'loss_rpn_cls': 0.083, time: 0.359
2019-08-28 10:23:40.920892, iter: 1200, lr: 0.01000, 'loss_cls': 0.527, 'loss_mask': 0.392, 'loss_bbox': 0.352, 'loss': 1.394, 'loss_rpn_bbox': 0.066, 'loss_rpn_cls': 0.097, time: 0.361
2019-08-28 10:23:59.175570, iter: 1250, lr: 0.01000, 'loss_cls': 0.556, 'loss_mask': 0.415, 'loss_bbox': 0.347, 'loss': 1.466, 'loss_rpn_bbox': 0.049, 'loss_rpn_cls': 0.078, time: 0.359
2019-08-28 10:24:17.054893, iter: 1300, lr: 0.01000, 'loss_cls': 0.482, 'loss_mask': 0.371, 'loss_bbox': 0.355, 'loss': 1.29, 'loss_rpn_bbox': 0.051, 'loss_rpn_cls': 0.063, time: 0.545
2019-08-28 10:24:35.260912, iter: 1350, lr: 0.01000, 'loss_cls': 0.396, 'loss_mask': 0.348, 'loss_bbox': 0.287, 'loss': 1.127, 'loss_rpn_bbox': 0.059, 'loss_rpn_cls': 0.051, time: 0.369
2019-08-28 10:24:53.680534, iter: 1400, lr: 0.01000, 'loss_cls': 0.447, 'loss_mask': 0.35, 'loss_bbox': 0.295, 'loss': 1.205, 'loss_rpn_bbox': 0.068, 'loss_rpn_cls': 0.064, time: 0.381
2019-08-28 10:25:11.668375, iter: 1450, lr: 0.01000, 'loss_cls': 0.614, 'loss_mask': 0.402, 'loss_bbox': 0.348, 'loss': 1.486, 'loss_rpn_bbox': 0.046, 'loss_rpn_cls': 0.081, time: 0.362
2019-08-28 10:25:29.691196, iter: 1500, lr: 0.01000, 'loss_cls': 0.531, 'loss_mask': 0.358, 'loss_bbox': 0.302, 'loss': 1.287, 'loss_rpn_bbox': 0.065, 'loss_rpn_cls': 0.053, time: 0.352
2019-08-28 10:25:48.003211, iter: 1550, lr: 0.01000, 'loss_cls': 0.394, 'loss_mask': 0.329, 'loss_bbox': 0.339, 'loss': 1.168, 'loss_rpn_bbox': 0.069, 'loss_rpn_cls': 0.056, time: 0.351
2019-08-28 10:26:06.506278, iter: 1600, lr: 0.01000, 'loss_cls': 0.313, 'loss_mask': 0.303, 'loss_bbox': 0.27, 'loss': 0.953, 'loss_rpn_bbox': 0.058, 'loss_rpn_cls': 0.041, time: 0.348
2019-08-28 10:26:24.911350, iter: 1650, lr: 0.01000, 'loss_cls': 0.358, 'loss_mask': 0.319, 'loss_bbox': 0.332, 'loss': 1.089, 'loss_rpn_bbox': 0.059, 'loss_rpn_cls': 0.041, time: 0.361
2019-08-28 10:26:43.280922, iter: 1700, lr: 0.01000, 'loss_cls': 0.343, 'loss_mask': 0.297, 'loss_bbox': 0.267, 'loss': 1.029, 'loss_rpn_bbox': 0.043, 'loss_rpn_cls': 0.04, time: 0.365
2019-08-28 10:27:01.706950, iter: 1750, lr: 0.01000, 'loss_cls': 0.257, 'loss_mask': 0.249, 'loss_bbox': 0.183, 'loss': 0.818, 'loss_rpn_bbox': 0.034, 'loss_rpn_cls': 0.027, time: 0.370
2019-08-28 10:27:20.418963, iter: 1800, lr: 0.01000, 'loss_cls': 0.385, 'loss_mask': 0.317, 'loss_bbox': 0.305, 'loss': 1.098, 'loss_rpn_bbox': 0.044, 'loss_rpn_cls': 0.045, time: 0.396
2019-08-28 10:27:38.672954, iter: 1850, lr: 0.01000, 'loss_cls': 0.313, 'loss_mask': 0.318, 'loss_bbox': 0.237, 'loss': 0.931, 'loss_rpn_bbox': 0.042, 'loss_rpn_cls': 0.032, time: 0.375
2019-08-28 10:27:56.875251, iter: 1900, lr: 0.01000, 'loss_cls': 0.304, 'loss_mask': 0.306, 'loss_bbox': 0.274, 'loss': 0.902, 'loss_rpn_bbox': 0.062, 'loss_rpn_cls': 0.044, time: 0.364
2019-08-28 10:28:15.098122, iter: 1950, lr: 0.01000, 'loss_cls': 0.386, 'loss_mask': 0.291, 'loss_bbox': 0.294, 'loss': 1.03, 'loss_rpn_bbox': 0.041, 'loss_rpn_cls': 0.035, time: 0.364
2019-08-28 10:28:33.250704, iter: 2000, lr: 0.01000, 'loss_cls': 0.312, 'loss_mask': 0.28, 'loss_bbox': 0.259, 'loss': 0.941, 'loss_rpn_bbox': 0.042, 'loss_rpn_cls': 0.025, time: 0.337
2019-08-28 10:28:51.688948, iter: 2050, lr: 0.01000, 'loss_cls': 0.226, 'loss_mask': 0.244, 'loss_bbox': 0.215, 'loss': 0.726, 'loss_rpn_bbox': 0.03, 'loss_rpn_cls': 0.024, time: 0.343
2019-08-28 10:29:10.119525, iter: 2100, lr: 0.01000, 'loss_cls': 0.233, 'loss_mask': 0.275, 'loss_bbox': 0.283, 'loss': 0.885, 'loss_rpn_bbox': 0.042, 'loss_rpn_cls': 0.032, time: 0.364
2019-08-28 10:29:28.423118, iter: 2150, lr: 0.01000, 'loss_cls': 0.216, 'loss_mask': 0.254, 'loss_bbox': 0.208, 'loss': 0.714, 'loss_rpn_bbox': 0.034, 'loss_rpn_cls': 0.025, time: 0.366
2019-08-28 10:29:47.098226, iter: 2200, lr: 0.01000, 'loss_cls': 0.273, 'loss_mask': 0.278, 'loss_bbox': 0.215, 'loss': 0.84, 'loss_rpn_bbox': 0.058, 'loss_rpn_cls': 0.032, time: 0.343
2019-08-28 10:30:05.369391, iter: 2250, lr: 0.01000, 'loss_cls': 0.237, 'loss_mask': 0.221, 'loss_bbox': 0.221, 'loss': 0.737, 'loss_rpn_bbox': 0.028, 'loss_rpn_cls': 0.02, time: 0.352
2019-08-28 10:30:23.709169, iter: 2300, lr: 0.01000, 'loss_cls': 0.301, 'loss_mask': 0.255, 'loss_bbox': 0.253, 'loss': 0.891, 'loss_rpn_bbox': 0.032, 'loss_rpn_cls': 0.024, time: 0.387
2019-08-28 10:30:41.909947, iter: 2350, lr: 0.01000, 'loss_cls': 0.234, 'loss_mask': 0.256, 'loss_bbox': 0.15, 'loss': 0.627, 'loss_rpn_bbox': 0.035, 'loss_rpn_cls': 0.017, time: 0.373
2019-08-28 10:30:59.774337, iter: 2400, lr: 0.01000, 'loss_cls': 0.175, 'loss_mask': 0.216, 'loss_bbox': 0.198, 'loss': 0.677, 'loss_rpn_bbox': 0.03, 'loss_rpn_cls': 0.023, time: 0.368
2019-08-28 10:31:17.912704, iter: 2450, lr: 0.01000, 'loss_cls': 0.205, 'loss_mask': 0.215, 'loss_bbox': 0.205, 'loss': 0.684, 'loss_rpn_bbox': 0.025, 'loss_rpn_cls': 0.018, time: 0.386
2019-08-28 10:31:36.230056, iter: 2500, lr: 0.01000, 'loss_cls': 0.182, 'loss_mask': 0.23, 'loss_bbox': 0.201, 'loss': 0.656, 'loss_rpn_bbox': 0.025, 'loss_rpn_cls': 0.018, time: 0.374
2019-08-28 10:31:54.746713, iter: 2550, lr: 0.01000, 'loss_cls': 0.163, 'loss_mask': 0.217, 'loss_bbox': 0.165, 'loss': 0.604, 'loss_rpn_bbox': 0.034, 'loss_rpn_cls': 0.015, time: 0.337
2019-08-28 10:32:13.348275, iter: 2600, lr: 0.01000, 'loss_cls': 0.139, 'loss_mask': 0.197, 'loss_bbox': 0.154, 'loss': 0.556, 'loss_rpn_bbox': 0.035, 'loss_rpn_cls': 0.02, time: 0.361
2019-08-28 10:32:31.558732, iter: 2650, lr: 0.01000, 'loss_cls': 0.188, 'loss_mask': 0.23, 'loss_bbox': 0.176, 'loss': 0.669, 'loss_rpn_bbox': 0.039, 'loss_rpn_cls': 0.026, time: 0.327
2019-08-28 10:32:49.654740, iter: 2700, lr: 0.01000, 'loss_cls': 0.193, 'loss_mask': 0.261, 'loss_bbox': 0.239, 'loss': 0.787, 'loss_rpn_bbox': 0.031, 'loss_rpn_cls': 0.031, time: 0.336
2019-08-28 10:33:25.747055, iter: 2800, lr: 0.01000, 'loss_cls': 0.155, 'loss_mask': 0.218, 'loss_bbox': 0.178, 'loss': 0.626, 'loss_rpn_bbox': 0.032, 'loss_rpn_cls': 0.026, time: 0.380
2019-08-28 10:33:44.111608, iter: 2850, lr: 0.01000, 'loss_cls': 0.125, 'loss_mask': 0.225, 'loss_bbox': 0.128, 'loss': 0.508, 'loss_rpn_bbox': 0.023, 'loss_rpn_cls': 0.017, time: 0.361
2019-08-28 10:34:02.944465, iter: 2900, lr: 0.01000, 'loss_cls': 0.143, 'loss_mask': 0.2, 'loss_bbox': 0.171, 'loss': 0.563, 'loss_rpn_bbox': 0.049, 'loss_rpn_cls': 0.015, time: 0.374
2019-08-28 10:34:21.538064, iter: 2950, lr: 0.01000, 'loss_cls': 0.114, 'loss_mask': 0.184, 'loss_bbox': 0.111, 'loss': 0.481, 'loss_rpn_bbox': 0.026, 'loss_rpn_cls': 0.013, time: 0.391

#固化模型
!cd mask-rcnn && python freeze_model.py 
                   --pretrained_model=output/model_final 
                   --freeze_model_save_dir=freeze_model 
                   --use_gpu=1

The COCO dataset or COCO API is not exist, use the default mapping of class index and real category name on COCO17.
W0828 10:36:57.041860 304 http://device_context.cc:259] Please NOTE: device: 0, CUDA Capability: 70, Driver API Version: 9.2, Runtime API Version: 9.0
W0828 10:36:57.046031 304 http://device_context.cc:267] device: 0, cuDNN Version: 7.3.
freeze succeed

#利用固化的模型进行预测
!cd mask-rcnn && python freeze_infer.py 
                  --image_path=coco30/train2017/000000005802.jpg  
                  --draw_threshold=0.7 
                  --use_gpu=1

######检测结果可视化
%matplotlib inline   
import matplotlib.pyplot as plt  
import numpy as np
import cv2

detect_img= cv2.imread('mask-rcnn/000000005802.jpg')

plt.imshow(detect_img)
plt.show()

The COCO dataset or COCO API is not exist, use the default mapping of class index and real category name on COCO17.
W0828 10:37:06.034216 373 http://device_context.cc:259] Please NOTE: device: 0, CUDA Capability: 70, Driver API Version: 9.2, Runtime API Version: 9.0
W0828 10:37:06.038113 373 http://device_context.cc:267] device: 0, cuDNN Version: 7.3.
image with bbox drawed saved as 000000005802.jpg