Super-resolution of multiphase materials by combining complementary 2D and 3D image data using generative adversarial networks

Abstract

Modelling the impact of a material's mesostructure on device level performance typically requires access to 3D image data containing all the relevant information to define the geometry of the simulation domain. This image data must include sufficient contrast between phases to distinguish each material, be of high enough resolution to capture the key details, but also have a large enough field-of-view to be representative of the material in general. It is rarely possible to obtain data with all of these properties from a single imaging technique. In this paper, we present a method for combining information from pairs of distinct but complementary imaging techniques in order to accurately reconstruct the desired multi-phase, high resolution, representative, 3D images. Specifically, we use deep convolutional generative adversarial networks to implement super-resolution, style transfer and dimensionality expansion. To demonstrate the widespread applicability of this tool, two pairs of datasets are used to validate the quality of the volumes generated by fusing the information from paired imaging techniques. Three key mesostructural metrics are calculated in each case to show the accuracy of this method. Having confidence in the accuracy of our method, we then demonstrate its power by applying to a real data pair from a lithium ion battery electrode, where the required 3D high resolution image data is not available anywhere in the literature. We believe this approach is superior to previously reported statistical material reconstruction methods both in terms of its fidelity and ease of use. Furthermore, much of the data required to train this algorithm already exists in the literature, waiting to be combined. As such, our open-access code could precipitate a step change in the computational materials sciences by generating the hard to obtain high quality image volumes necessary to simulate behaviour at the mesoscale.

Machine-Learning-Driven Advanced Characterization of Battery Electrodes

“…In each example, the image on the left shows the input, and the image on the right shows the output after the specified method was applied. In each case, the images were generated based on the approaches described in relevant studies: segmentation, 26 generation, 41 inpainting, 42 style transfer, 40 superresolution, 43 and dimensionality expansion. Missing or corrupted regions of an image can be filled in with a technique called inpainting.…”

“…There have been great advancements in deep learning SR, most notably, for the application of facial recognition by super-resolving low-res faces taken in the natural environment. These ML-based methods employ CNNs with a range of architectures from GANs 40,60 to deep residual nets. 61 It is important to note the difference between superresolution and super-sampling in the context of imaging for electrode characterization.…”

“…As such, we are especially interested in the new family of machine learning (ML) techniques that can achieve greater breadth of application, with comparatively little expert finetuning required. For example, Dahari et al 40 used data from distinct but complementary imaging techniques (XCT and SEM) to train deep convolution generative adversarial networks (GANs) to generate representative 3D images with multi-modal details. The opportunities to expand beyond this work to create high-dimensional data fusion and superresolution datasets will be discussed in the following section.…”

“…Aside from optimization, the ML methods described in the previous section also offer a new approach for obtaining multimodal datasets through data fusion, as was recently explored by Dahari et al 40 As an example workflow, starting from a high-resolution SEM image, segmentation could first be used to generate a phase map. Dimensionality expansion can then be applied to generate a large 3D volume.…”

“…Examples of machine learning methods for enabling next-generation materials characterization.In each example, the image on the left shows the input, and the image on the right shows the output after the specified method was applied. In each case, the images were generated based on the approaches described in relevant studies: segmentation,26 generation,41 inpainting, 42 style transfer,40 superresolution, 43 and dimensionality expansion 44.…”