Interactive Visual Study of Multiple Attributes Learning Model of X-Ray Scattering Images


Abstract

Existing interactive visualization tools for deep learning are mostly applied to the training, debugging, and refinement of neural network models working on natural images. While deep learning methods are applied to scientific images of x-ray scattering, the visual analysis should focus on the classification behavior of multiple structural attributes, which however, has not been well supported by existing visualization tools. In this paper, we present an interactive system for domain scientists to visually study the multiple attributes learning models applied to x-ray scattering images. It allows domain scientists to explore the scientific images in the embedded spaces of the model prediction output, the actual labels, and the discovered feature space of neural networks. The interactions allow users to flexibly select instance images, their clusters, and compare them with specific visual representation of attributes. The exploration is guided by the manifestation of model performance related to mutual relationships among attributes, which often affect the learning accuracy and effectiveness. The system thus allows domain scientists to improve the training dataset and model, find questionable attributes labels, and identify outlier images or spurious data clusters. Preliminary case studies and scientists feedback demonstrate its functionalities and usefulness.


The online version of X-Scattering is accessible HERE

Introduction


Case Study 1: Studying model performance from an ACT image group


Case Study 2: Studying model behavior with three co-existence attributes