Recovering_3D_Basement_Relief_Using_Gravity_Data_Through_Convolutional_Neural_Networks folder includes a novel method using the convolutional neural network (CNN) for depth-to-basement inversion directly from gravity data. 

Here is the directory structure of the folder:

-------------Recovering_3D_Basement_Relief_Using_Gravity_Data_Through_Convolutional_Neural_Networks-----------
Synthetic_examples
	--Symmetric_model
	--Asymmetric_model
	--Complex_model
Field_example

XXX_model or Field_example generally includes:

	code:
		Training_XXX_model.py // train CNN
		Predict_XXX_model.py // recover basement relief
	data:
		XXX_dataset // data set for training CNN
		XXX_model_forward // observation data and the parameters of density contrast
		Recovered_basement_XXX // inversion result
		XXX.h5 // trained network
		XXX.txt // the loss of trained network
-------------Recovering_3D_Basement_Relief_Using_Gravity_Data_Through_Convolutional_Neural_Networks-----------

Main data and parameters are described as follow:
-------------------------Main data and parameters-------------------
XXX_model_forward.mat or field_case_data.mat generally contains:

aa, bb, cc, dd // the parameters of density contrast.
gam // gravitational constant
H0 // the height of the reference plane
real_obs // Matrix form of observed data
rho_aver // the mean density contrast used for Bouguer model estimation
data // observed data and their position. 
The first column is xp, the second column is yp, the third column is zp, the fourth column is comp (component for the gravity data, the value of 3 represents gz component which is the default value for this code), the fifth column is the gravity data.
X Y Z is the true model. 
-------------------------Main data and parameters-------------------

-------------------------------------------------------------------------------------
Performing an inversion as follow steps:
1) Enter XXX_model or Field_example folder.
2) using Training_XXX_model.py to get the trained network (XXX.h5) and the corresponding loss (XXXloss.txt).
3) Then, we load the trained network and observed data to recover the basement relief using predict_XXX_model.py
-------------------------------------------------------------------------------------

--------------------------------Requirement------------------------------
	python 3.6
		tensorflow 2.1.1
		tensorflow-gpu 2.1.0
		h5py 2.9.0
		matplotlib 3.0.3
		scipy 1.4.1
		numpy 1.16.2
--------------------------------Requirement------------------------------

----------------------COPYRIGHTS---------------------------------------------
These programs may be freely redistributed under the condition that the 
copyright notices are not removed, and no compensation is received.  Private, 
research, and institutional use is free. You may distribute modified versions 
of this code UNDER THE CONDITION THAT THIS CODE AND ANY MODIFICATIONS MADE TO 
IT IN THE SAME FILE REMAIN UNDER COPYRIGHT OF THE ORIGINAL AUTHOR, BOTH SOURCE 
AND OBJECT CODE ARE MADE FREELY AVAILABLE WITHOUT CHARGE, AND CLEAR NOTICE IS 
GIVEN OF THE MODIFICATIONS. Distribution of this code as part of a commercial 
system is permissible ONLY BY DIRECT ARRANGEMENT WITH THE AUTHOR. If you use the code and data
 in the Recovering_3D_Basement_Relief_Using_Gravity_Data_Through_Convolutional_Neural_Networks folder, 
and especially if you use it to accomplish real work, PLEASE SEND ME AN EMAIL. 

Copyright 2021, Hongzhu Cai
2021-10-01
-----------------------------------------------------------------------------

For possible errors, please contact:

     caihongzhu@hotmail.com
     Hongzhu Cai
     China University of Geosciences, Wuhan, 
     Hubei, China.
     2021-10-01