A programm package based on deeplabv3+ neural network for statistical characteristics of gas-liquid two-phase flow written in python.
The parameters used in the code are based on the experimental results of semantic segmentation of gas-liquid two-phase flow using the deeplabv3+ neural network.
The link of source code of deeplabv3+ I used：
deeplabv3+
The link of the weight file:
Weight file
Remember to modify the parameters in the neural network,
name_classes = ["background","liquid","gas"]
where gas corresponds to green color, and liquid corresponds to red color.
You can configure the environment required to run the program through the following commands.
pip install -r requirements.txt
In file general.py:
#---------Get length and area distribution diagram---------#
len_sqr=1
#-----------Get the speed distribution map ------------#
velocity=1
#----------Whether it is a variable cross-section microchannel -----------#
#Constant cross-section microchannel-mode=0 Variable cross-section microchannel-mode=1#
mode=0
#---------Path where stores the video-------#
path='C:/GraduateWork/deeplabv3-plus-pytorch/general/video/'
#---------The name of video file(for example: 3_liq_3_gas_mid_CR600x2_1836-ST-B-062_1.avi)----------#
path_avi=os.path.join(path,'3_liq_3_gas_mid_CR600x2_1836-ST-B-062_1.avi')
#------------------------------------------------------------------------------------------------------------------------#
#---------Change the path C:/GraduateWork/deeplabv3-plus-pytorch/predict.py into yourselfs-------#
res=os.popen('python C:/GraduateWork/deeplabv3-plus-pytorch/predict.py --pathi %s --patho %s'%(path_png,path_data))
If you have followed the instructions above, please just run general.py directly.
Example of the original video of gas-liquid two-phase flow:
Example of the processed video of gas-liquid two-phase flow:
Distribution diagram of length and area of gas-liquid two-phase flow gived by programm:
Distribution diagram of veolocity of gas-liquid two-phase flow gived by programm:
