search

xinwucwp / faultSeg

Using synthetic datasets to train an end-to-end CNN for 3D fault segmentation (We are working on an improved version!)
206 stars 130 forks source link

Modifying the model structure #12

Closed 1197155807 closed 2 years ago

1197155807 commented 3 years ago

Dear Dr. Wu thank you for your great effort。 If I want to modify the model model/fseg-60.hdf5 file and I want to use a new structure as my convolutional neural network, how should I modify it? Thanks again.

llmpass commented 3 years ago

You can simply write/design a new one from scratch, I see it's better and way simpler than modifying from the existing hdf5 file.

1197155807 commented 3 years ago

First of all, thank you very much for your answer. If I want to build my own HDF5 model, is there any way that I could see the specific code in the file FSEG-60. HDF5? I want to know, I want to know, what package is it built from. Thank you again!

llmpass commented 3 years ago

It's created in train.py: https://github.com/xinwucwp/faultSeg/blob/40a3bae2b7980916b7f3091ca058e363b05696fc/train.py#L44 You can go modify unet3.py to create a new model and change the file name in train.py accordingly. The package is keras. Hope this helps.