3D-Atom-Segmentation

This repository implements 3D convolutional neural network augmentation of atomic resolution electron tomograms using UNet architectures. The two modified artictectures used in this package are 1) Accurate and versatile 3D segmentation of plant tissues at cellular resolution - Adrian Wolny et al and 2) R2U3D: Recurrent Residual 3D U-Net for Lung Segmentation - Dhaval Kadia.

This package uses experimental reconstructions and labeled coordinates from data available through the following publications:

1. Deciphering chemical order/disorder and material properties at the single-atom level
2. Observing crystal nucleation in four dimensions using atomic electron tomography
3. Determining the three-dimensional atomic structure of an amorphous solid
4. Three-dimensional atomic packing in amorphous solids with liquid-like structure
5. Simultaneous Successive Twinning Captured by Atomic Electron Tomography

If you have any questions, feel free to contact - Jonathan Schwartz: [email protected]