search

RhFeng / UN_PORO

Unsupervised learning reservoir porosity
1 stars 0 forks source link

Porosity and AI relation, amount of filters in the second layer of CNN, data type #1

Open toghrul97 opened 2 years ago

toghrul97 commented 2 years ago

Hello, I have following questions about the project. 1.The equation describing AI and porosity is p = 0.02AI − 0.06. But in the code it is 0.2AI - 0.06

  1. Also on the paper it was written that the second layer of CNN is composed of 24 filters, whereas in the code it is 12 in both layers.
  2. Can I use this model by giving an input as seismic facies in GSLIB format? The data is in the following link https://drive.google.com/file/d/1L-o3mgmt-bo-yen3HodpvjBdoqVXGmot/view?usp=sharing
RhFeng commented 2 years ago
  1. It depends on the scale of AI that is used. If the scale of AI is between [0 1], then it should be 0.2, if it is between [0 10], 0.02 is used.
  2. I didn't quite check the network structure with the one used in the paper. But the point here is for people to play around and see what the result gonna be. For example the error function used in the code is different from the one mentioned in the paper.
  3. The input data should be in ascii format. However, other format is accepted, if it can be transferred as array in python.