Object grasping in cluttered scenes is a widely investigated field of robot manipulation. Most of the current works focus on estimating grasp pose from point clouds based on an efficient single-shot grasp detection network. However, due to the lack o
f geometry awareness of the local grasping area, it may cause severe collisions and unstable grasp configurations. In this paper, we propose a two-stage grasp pose refinement network which detects grasps globally while fine-tuning low-quality grasps and filtering noisy grasps locally. Furthermore, we extend the 6-DoF grasp with an extra dimension as grasp width which is critical for collisionless grasping in cluttered scenes. It takes a single-view point cloud as input and predicts dense and precise grasp configurations. To enhance the generalization ability, we build a synthetic single-object grasp dataset including 150 commodities of various shapes, and a multi-object cluttered scene dataset including 100k point clouds with robust, dense grasp poses and mask annotations. Experiments conducted on Yumi IRB-1400 Robot demonstrate that the model trained on our dataset performs well in real environments and outperforms previous methods by a large margin.
We study solutions of the Bethe ansatz equations associated to the orthosymplectic Lie superalgebras $mathfrak{osp}_{2m+1|2n}$ and $mathfrak{osp}_{2m|2n}$. Given a solution, we define a reproduction procedure and use it to construct a family of new s
olutions which we call a population. To each population we associate a symmetric rational pseudo-differential operator $mathcal R$. Under some technical assumptions, we show that the superkernel $W$ of $mathcal R$ is a self-dual superspace of rational functions, and the population is in a canonical bijection with the variety of isotropic full superflags in $W$ and with the set of symmetric complete factorizations of $mathcal R$. In particular, our results apply to the case of even Lie algebras of type D${}_m$ corresponding to $mathfrak{osp}_{2m|0}=mathfrak{so}_{2m}$.
We give explicit actions of Drinfeld generators on Gelfand-Tsetlin bases of super Yangian modules associated with skew Young diagrams. In particular, we give another proof that these representations are irreducible. We study irreducible tame $mathrm
Y(mathfrak{gl}_{1|1})$-modules and show that a finite-dimensional irreducible $mathrm Y(mathfrak{gl}_{1|1})$-module is tame if and only if it is thin. We also give the analogous statements for quantum affine superalgebra of type A.
Despite the vast literature on Human Activity Recognition (HAR) with wearable inertial sensor data, it is perhaps surprising that there are few studies investigating semisupervised learning for HAR, particularly in a challenging scenario with class i
mbalance problem. In this work, we present a new benchmark, called A*HAR, towards semisupervised learning for class-imbalanced HAR. We evaluate state-of-the-art semi-supervised learning method on A*HAR, by combining Mean Teacher and Convolutional Neural Network. Interestingly, we find that Mean Teacher boosts the overall performance when training the classifier with fewer labelled samples and a large amount of unlabeled samples, but the classifier falls short in handling unbalanced activities. These findings lead to an interesting open problem, i.e., development of semi-supervised HAR algorithms that are class-imbalance aware without any prior knowledge on the class distribution for unlabeled samples. The dataset and benchmark evaluation are released at https://github.com/I2RDL2/ASTAR-HAR for future research.
Magnetic skyrmions are well-suited for encoding information because they are nano-sized, topologically stable, and only require ultra-low critical current densities $j_c$ to depin from the underlying atomic lattice. Above $j_c$ skyrmions exhibit well
-controlled motion, making them prime candidates for race-track memories. In thin films thermally-activated creep motion of isolated skyrmions was observed below $j_c$ as predicted by theory. Uncontrolled skyrmion motion is detrimental for race-track memories and is not fully understood. Notably, the creep of skyrmion lattices in bulk materials remains to be explored. Here we show using resonant ultrasound spectroscopy--a probe highly sensitive to the coupling between skyrmion and atomic lattices--that in the prototypical skyrmion lattice material MnSi depinning occurs at $j_c^*$ that is only 4 percent of $j_c$. Our experiments are in excellent agreement with Anderson-Kim theory for creep and allow us to reveal a new dynamic regime at ultra-low current densities characterized by thermally-activated skyrmion-lattice-creep with important consequences for applications.
We suggest the notion of perfect integrability for quantum spin chains and conjecture that quantum spin chains are perfectly integrable. We show the perfect integrability for Gaudin models associated to simple Lie algebras of all finite types, with p
eriodic and regular quasi-periodic boundary conditions.
We show that the quantum Berezinian which gives a generating function of the integrals of motions of XXX spin chains associated to super Yangian $mathrm{Y}(mathfrak{gl}_{m|n})$ can be written as a ratio of two difference operators of orders $m$ and $
n$ whose coefficients are ratios of transfer matrices corresponding to explicit skew Young diagrams. In the process, we develop several missing parts of the representation theory of $mathrm{Y}(mathfrak{gl}_{m|n})$ such as $q$-character theory, Jacobi-Trudi identity, Drinfeld functor, extended T-systems, Harish-Chandra map.
Deep learning (DL) based object detection has achieved great progress. These methods typically assume that large amount of labeled training data is available, and training and test data are drawn from an identical distribution. However, the two assum
ptions are not always hold in practice. Deep domain adaptive object detection (DDAOD) has emerged as a new learning paradigm to address the above mentioned challenges. This paper aims to review the state-of-the-art progress on deep domain adaptive object detection approaches. Firstly, we introduce briefly the basic concepts of deep domain adaptation. Secondly, the deep domain adaptive detectors are classified into five categories and detailed descriptions of representative methods in each category are provided. Finally, insights for future research trend are presented.
Object detection in thermal images is an important computer vision task and has many applications such as unmanned vehicles, robotics, surveillance and night vision. Deep learning based detectors have achieved major progress, which usually need large
amount of labelled training data. However, labelled data for object detection in thermal images is scarce and expensive to collect. How to take advantage of the large number labelled visible images and adapt them into thermal image domain, is expected to solve. This paper proposes an unsupervised image-generation enhanced adaptation method for object detection in thermal images. To reduce the gap between visible domain and thermal domain, the proposed method manages to generate simulated fake thermal images that are similar to the target images, and preserves the annotation information of the visible source domain. The image generation includes a CycleGAN based image-to-image translation and an intensity inversion transformation. Generated fake thermal images are used as renewed source domain. And then the off-the-shelf Domain Adaptive Faster RCNN is utilized to reduce the gap between generated intermediate domain and the thermal target domain. Experiments demonstrate the effectiveness and superiority of the proposed method.
We study the $mathfrak{gl}_{1|1}$ supersymmetric XXX spin chains. We give an explicit description of the algebra of Hamiltonians acting on any cyclic tensor products of polynomial evaluation $mathfrak{gl}_{1|1}$ Yangian modules. It follows that there
exists a bijection between common eigenvectors (up to proportionality) of the algebra of Hamiltonians and monic divisors of an explicit polynomial written in terms of the Drinfeld polynomials. In particular our result implies that each common eigenspace of the algebra of Hamiltonians has dimension one. We also give dimensions of the generalized eigenspaces. We show that when the tensor product is irreducible, then all eigenvectors can be constructed using Bethe ansatz. We express the transfer matrices associated to symmetrizers and anti-symmetrizers of vector representations in terms of the first transfer matrix and the center of the Yangian.
