Feature Normalization (FN) is an important technique to help neural network training, which typically normalizes features across spatial dimensions. Most previous image inpainting methods apply FN in their networks without considering the impact of the corrupted regions of the input image on normalization, e.g. mean and variance shifts. In this work, we show that the mean and variance shifts caused by full-spatial FN limit the image inpainting network training and we propose a spatial region-wise normalization named Region Normalization (RN) to overcome the limitation. RN divides spatial pixels into different regions according to the input mask, and computes the mean and variance in each region for normalization. We develop two kinds of RN for our image inpainting network: (1) Basic RN (RN-B), which normalizes pixels from the corrupted and uncorrupted regions separately based on the original inpainting mask to solve the mean and variance shift problem; (2) Learnable RN (RN-L), which automatically detects potentially corrupted and uncorrupted regions for separate normalization, and performs global affine transformation to enhance their fusion. We apply RN-B in the early layers and RN-L in the latter layers of the network respectively. Experiments show that our method outperforms current state-of-the-art methods quantitatively and qualitatively. We further generalize RN to other inpainting networks and achieve consistent performance improvements.

X-ray diffraction and ab initio molecular dynamics simulation studies of molten Al 60 Cu 40 have been carried out between 973 and 1323 K. The structures obtained from our simulated atomic models are fully consistent with the experimental results. The local structures of the models analyzed using Honeycutt-Andersen and Voronoi tessellation methods clearly demonstrate that as the temperatures of the liquid is lowered it becomes more ordered. While no one cluster-type dominates the local structure of this liquid, the most prevalent polyhedra in the liquid structure can be described as distorted icosahedra. No obvious correlations between the clusters observed in the liquid and known stable crystalline phases in this system were observed.

Abstract:Revealing the essential structural features of metallic glasses (MGs) will enhance the understanding of glass-forming mechanisms. In this work, a feasible scheme is provided where we performed the state-of-the-art synchrotron-radiation based experiments combined with simulations to investigate the microstructures of ZrCu amorphous compositions. It is revealed that in order to stabilize the amorphous state and optimize the topological and chemical distribution, besides the icosahedral or icosahedral-like clusters, other types of clusters also participate in the formation of the microstructure in MGs. This cluster-level co-existing feature may be popular in this class of glassy materials.

No abstract

No abstract