Search

Modern semantic segmentation frameworks usually combine low-level and high-level features from pre-trained backbone convolutional models to boost performance. In this paper, we first point out that a simple fusion of low-level and high-level features could be less effective because of the gap in semantic levels and spatial resolution. We find that introducing semantic information into low-level features and high-resolution details into high-level features is more effective for the later fusion. Based on this observation, we propose a new framework, named ExFuse, to bridge the gap between low-level and high-level features thus significantly improve the segmentation quality by 4.0\% in total. Furthermore, we evaluate our approach on the challenging PASCAL VOC 2012 segmentation benchmark and achieve 87.9\% mean IoU, which outperforms the previous state-of-the-art results.

Fabrication of diamond structures by self-assembly is a fundamental challenge in making three-dimensional photonic crystals. We simulate a system of model hard particles with attractive patches and show that they can self-assemble into a diamond structure from an initially disordered state. We quantify the extent to which the formation of the diamond structure can be facilitated by "seeding" the system with small diamond crystallites or by introducing a rotation interaction to mimic a carbon-carbon antibonding interaction. Our results suggest patchy particles may serve as colloidal "atoms" and "molecules" for the bottom-up self-assembly of three-dimensional crystals., Comment: 6 pages, 5 figures, 1 table; Langmuir 21, 25 (2005)

We systematically investigate the self-assembly of anisotropic cone-shaped particles decorated by ring-like attractive patches. We demonstrate that the self-assembled clusters, which arise due to the conical particle's anisotropic shape combined with directional attractive interactions, are precise for certain cluster sizes, resulting in a precise packing sequence of clusters of increasing sizes with decreasing cone angles. We thoroughly explore the dependence of cluster packing on cone angle and cooling rate, and categorize the resulting structures as stable and metastable clusters. We also discuss the implication of our simulation results in the context of the Israelachvili packing rule for surfactants, and a recent geometrical packing analysis on hard cones in the limit of large numbers of cones.

Molecular simulations of the self-assembly of cone-shaped particles with specific, attractive interactions are performed. Upon cooling from random initial conditions, we find that the cones self assemble into clusters and that clusters comprised of particular numbers of cones (e.g. 4 - 17, 20, 27, 32, 42) have a unique and precisely packed structure that is robust over a range of cone angles. These precise clusters form a sequence of structures at specific cluster sizes- a precise packing sequence - that for small sizes is identical to that observed in evaporation-driven assembly of colloidal spheres. We further show that this sequence is reproduced and extended in simulations of two simple models of spheres self-assembling from random initial conditions subject to certain convexity constraints. This sequence contains six of the most common virus capsid structures obtained in vivo including large chiral clusters, and a cluster that may correspond to several non-icosahedral, spherical virus capsid structures obtained in vivo. Our findings suggest this precise packing sequence results from free energy minimization subject to convexity constraints and is applicable to a broad range of assembly processes., Comment: 23 pages, 3 figures