We present a simple, fully-convolutional model for real-time (> 30 fps) instance segmentation that achieves competitive results on MS COCO evaluated on a single Titan Xp, which is significantly faster than any previous state-of-the-art approach. Moreover, we obtain this result after training on only one GPU. We accomplish this by breaking instance segmentation into two parallel subtasks: (1) generating a set of prototype masks and (2) predicting per-instance mask coefficients. Then we produce instance masks by linearly combining the prototypes with the mask coefficients. We find that because this process doesn't depend on repooling, this approach produces very high-quality masks and exhibits temporal stability for free. Furthermore, we analyze the emergent behavior of our prototypes and show they learn to localize instances on their own in a translation variant manner, despite being fully-convolutional. We also propose Fast NMS, a drop-in 12 ms faster replacement for standard NMS that only has a marginal performance penalty. Finally, by incorporating deformable convolutions into the backbone network, optimizing the prediction head with better anchor scales and aspect ratios, and adding a novel fast mask re-scoring branch, our YOLACT++ model can achieve 34.1 mAP on MS COCO at 33.5 fps, which is fairly close to the state-of-the-art approaches while still running at real-time.

We propose 'Hide-and-Seek', a weakly-supervised framework that aims to improve object localization in images and action localization in videos. Most existing weakly-supervised methods localize only the most discriminative parts of an object rather than all relevant parts, which leads to suboptimal performance. Our key idea is to hide patches in a training image randomly, forcing the network to seek other relevant parts when the most discriminative part is hidden. Our approach only needs to modify the input image and can work with any network designed for object localization. During testing, we do not need to hide any patches. Our Hide-and-Seek approach obtains superior performance compared to previous methods for weakly-supervised object localization on the ILSVRC dataset. We also demonstrate that our framework can be easily extended to weakly-supervised action localization.

We present an approach to discover and segment foreground object (s)

We present a novel deep neural network architecture for end-to-end scene flow estimation that directly operates on large-scale 3D point clouds. Inspired by Bilateral Convolutional Layers (BCL), we propose novel DownBCL, Up-BCL, and CorrBCL operations that restore structural information from unstructured point clouds, and fuse information from two consecutive point clouds. Operating on discrete and sparse permutohedral lattice points, our architectural design is parsimonious in computational cost. Our model can efficiently process a pair of point cloud frames at once with a maximum of 86K points per frame. Our approach achieves state-of-the-art performance on the Fly-ingThings3D and KITTI Scene Flow 2015 datasets. Moreover, trained on synthetic data, our approach shows great generalization ability on real-world data and on different point densities without fine-tuning.

Powder diffraction patterns of the zeolites natrolite (Na(16)Al(16)Si(24)O(80).16H(2)O), mesolite (Na(5.33)Ca(5.33)Al(16)Si(24)O(80).21.33H(2)O), scolecite (Ca(8)Al(16)Si(24)O(80).24H(2)O), and a gallosilicate analogue of natrolite (K(16)Ga(16)Si(24)O(80).12H(2)O), all crystallizing with a natrolite framework topology, were measured as a function of pressure up to 5.0 GPa with use of a diamond-anvil cell and a 200 microm focused monochromatic synchrotron X-ray beam. Under the hydrostatic conditions mediated by an alcohol and water mixture, all these materials showed an abrupt volume expansion (ca. 2.5% in natrolite) between 0.8 and 1.5 GPa without altering the framework topology. Rietveld refinements using the data collected on natrolite show that the anomalous swelling is due to the selective sorption of water from the pressure-transmission fluid expanding the channels along the a- and b-unit cell axes. This gives rise to a "superhydrated" phase of natrolite with an approximate formula of Na(16)Al(16)Si(24)O(80).32H(2)O, which contains hydrogen-bonded helical water nanotubes along the channels. In mesolite, which at ambient pressure is composed of ordered layers of sodium- and calcium-containing channels in a 1:2 ratio along the b-axis, this anomalous swelling is accompanied by a loss of the superlattice reflections (b(mesolite) = 3b(natrolite)). This suggests a pressure-induced order-disorder transition involving the motions of sodium and calcium cations either through cross-channel diffusion or within the respective channels. The powder diffraction data of scolecite, a monoclinic analogue of natrolite where all sodium cations are substituted by calcium and water molecules, reveal a reversible pressure-induced partial amorphization under hydrostatic conditions. Unlike the 2-dimensional swelling observed in natrolite and mesolite, the volume expansion of the potassium gallosilicate natrolite is 3-dimensional and includes the lengthening of the channel axis. In addition, the expanded phase, stable at high pressure, is retained at ambient conditions after pressure is released. The unprecedented and intriguing high-pressure crystal chemistry of zeolites with the natrolite framework topology is discussed here relating the different types of volume expansion to superhydration.

The thermal expansion and heat capacity of FeSb2 at ambient pressure agrees with a picture of a temperature induced spin state transition within the Fe t2g multiplet. However, high pressure powder diffraction data show no sign of a structural phase transition up to 7GPa. A bulk modulus B=84(3)GPa has been extracted and the temperature dependence of the Grüneisen parameter has been determined. We discuss here the relevance of a Kondo insulator description for this material.

No abstract

The effect of structure-directing agents (SDAs) on zeolite particle morphology and crystal structure was studied. Dimers and trimers of tetrapropylammonium hydroxidesthe SDA of siliceous ZSM-5 (silicalite-1, structure type MFI)slead to direct growth of the same framework but with drastically different crystal morphologies. Control over relative crystal growth rates along the three main crystallographic axes is demonstrated by obtaining crystals with their longest dimension along any of these axes. Unit cell dimensions and crystal structure were determined by refinement based on powder X-ray diffraction data and electron microscopy, respectively. The importance of the new morphologies in thin film molecular sieve technology along with possible crystal growth mechanisms is discussed.