nu-cucis/localizationdl

A Deep Learning Model For Localization of Two-Phase High-Contrast Three-dimensional Materials

Deep learning for localization of two phase high contrast three-dimensional material

LocalizationDL is an deep learning application for modeling localization linkages for two phase high contrast three-dimensional material. It’s a feature-engineering-free framework, which directly takes raw data as input, and trains a convolutional neural network (CNN).

To use this software, what the algorithm requires as input are a numpy array. The shape of this numpy array is (x, 11, 11, 11) where x is the number of focal voxels and the dimension of microstructure centered at focal voxels should be three-dimensional (i.e. 11x11x11). The software will take the row data as input, and train the predictive model. The detailed drscription about data preprocessing and model can be found in the published paper given below.

Requirements

• Python 3.6.3;
• Numpy 1.18.1;
• Sklearn 0.20.0;
• Keras 2.3.1;
• Pickle 4.0;
• TensorFlow 2.1.0;
• h5py 2.9.0;

Files

1. contrast10_localization.py: The script to train CNN for contrast 10 dataset, and save the model in 'my_model.h5' file.
2. contrast50_localization.py: The script to train CNN for contrast 50 dataset, and save the model in 'my_model.h5' file.
3. sample_data.pkl: Example data for contrast10 dataset, including 10 21x21x21 3D microstructure.
4. sample_data50.pkl: Example data for contrast50 dataset, including 10 21x21x21 3D microstructure.

How to run it

1. Run commend below, which trains the CNN model for contrast 10 dataset and save the trained model in 'my_model.h5' file.

   python contrast10_localization.py
   

2. Run commend below, which trains the CNN model for contrast 50 dataset and save the trained model in 'my_model.h5' file.

   python contrast50_localization.py
   

Acknowledgement

This work is supported in part by the following grants: AFOSR award FA9550-12-1-0458; NIST award 70NANB14H012; NSF award CCF-1409601; DOE awards DESC0007456, DE-SC0014330; and Northwestern Data Science Initiative.

Related Publications

Z. Yang, Y. C. Yabansu, D. Jha, W.-keng Liao, A. N. Choudhary, S. R. Kalidindi, and A. Agrawal, “Establishing structure-property localization linkages for elastic deformation of three-dimensional high contrast composites using deep learning approaches,” Acta Materialia, vol. 166, pp. 335–345, 2019.

Contact

Zijiang Yang ([email protected]); Ankit Agrawal ([email protected]); Alok Choudhary ([email protected]);