RESUMO
The problem of the efficient design of material microstructures exhibiting desired properties spans a variety of engineering and science applications. The ability to rapidly generate microstructures that exhibit user-specified property distributions can transform the iterative process of traditional microstructure-sensitive design. We reformulate the microstructure design process using a constrained generative adversarial network (GAN) model. This approach explicitly encodes invariance constraints within GANs to generate two-phase morphologies for photovoltaic applications obeying design specifications: specifically, user-defined short-circuit current density and fill factor combinations. Such invariance constraints can be represented by differentiable, deep learning-based surrogates of full physics models mapping microstructures to photovoltaic properties. Furthermore, we propose a multi-fidelity surrogate that reduces expensive label requirements by a factor of five. Our framework enables the incorporation of expensive or non-differentiable constraints for the fast generation of microstructures (in 190 ms) with user-defined properties. Such proposed physics-aware data-driven methods for inverse design problems can be used to considerably accelerate the field of microstructure-sensitive design.
RESUMO
This document describes the collection and organization of a dataset that consists of raw videos and extracted sub-images from video frames of a promising additive manufacturing technique called Two-Photon Lithography (TPL). Four unprocessed videos were collected, with each video capturing the printing process of different combinations of 3D parts on different photoresists at varying light dosages. These videos were further trimmed to obtain short clips that are organized by experimental parameters. Additionally, this dataset also contains a python script to reproduce an organized directory of cropped video frames extracted from the trimmed videos. These cropped frames focus on a region of interest around the parts being printed. We envision that the raw videos and cropped frames provided in this dataset will be used to train various computer vision and machine learning algorithms for applications such as object segmentation and localization applications. The cropped video frames were manually labelled by an expert to determine the quality of the printed part and for printing parameter optimization as presented in "Automated Detection of Part Quality during Two-Photon Lithography via Deep Learning" [1].
