In this paper, we propose a self-critical attention learning method for person re-identification. Unlike most existing methods which train the mechanism in weakly-supervised manner and ignore confidence level, learn with critic measures quality provides powerful supervisory signal to guide process. Moreover, model facilitates interpretation of effectiveness during process, by estimating maps. Specifically, jointly our agent reinforcement manner, where produces visual while analyzes gain from...

In this paper, we propose a consistent-aware deep learning (CADL) framework for person re-identification in camera network. Unlike most existing methods which identify whether two body images are from the same person, our approach aims to obtain maximal correct matches whole Different recently proposed network based only consider consistent information matching stage global optimal association, exploit such under where both feature representation and image automatically learned with certain...

In this paper, we propose a uniform and variational deep learning (UVDL) method for RGB-D object recognition person re-identification. Unlike most existing re-identification methods, which usually use only the visual appearance information from RGB images, our recognizes objects persons with images to exploit more reliable such as geometric anthropometric that are robust different viewpoints. Specifically, extract depth feature two convolutional neural networks, respectively. order combine...

Trajectory prediction is confronted with the dilemma to capture multi-modal nature of future dynamics both diversity and accuracy. In this paper, we present a distribution discrimination (DisDis) method predict personalized motion patterns by distinguishing potential distributions. Motivated that pattern each person due his/her habit, our DisDis learns latent represent different optimize it contrastive discrimination. This encourages distributions be more discriminative. Our can integrated...

In response to the complex unstructured environment of coal mines, a design scheme for variable wheel diameter robots is proposed. Based on gear, connecting rod and sliding rail mechanism, with was designed. Through change in robot, obstacle-crossing ability terrain adaptation robot were improved. The kinematic model single established variation rule radial length analysed. kinematics whole vehicle established, motion state under different driving speeds RecurDyn software, turning, process...

The point cloud based 3D single object tracking has drawn increasing attention. Although many breakthroughs have been achieved, we also reveal two severe issues. By extensive analysis, find the prediction manner of current approaches is non-robust, i.e., exposing a misalignment gap between score and actually localization accuracy. Another issue sparse returns will damage feature matching procedure SOT task. Based on these insights, introduce novel modules, Adaptive Refine Prediction (ARP)...

In this paper, we propose an Enhanced Bayesian Compression method to flexibly compress the deep networks via reinforcement learning. Unlike existing compression which cannot explicitly enforce quantization weights during training, our learns flexible codebooks in each layer for optimal network quantization. To dynamically adjust state of codebooks, employ Actor-Critic collaborate with original network. Different from most methods, EBC does not require re-training procedures after...

Robotic manipulation, owing to its multi-modal nature, often faces significant training ambiguity, necessitating explicit instructions clearly delineate the manipulation details in tasks. In this work, we highlight that vision instruction is naturally more comprehensible recent robotic policies than commonly adopted text instruction, as these are born with some understanding ability like human infants. Building on premise and drawing inspiration from cognitive science, introduce imagery...

Abstract Double‐vacuum‐bag (DVB) process, as an effective out‐of‐autoclave (OoA) technology, can manufacture composite materials with low void content and high mechanical properties. Therefore, this paper aimed to achieve content, the effects of pressure difference between inner outer bags, dwelling temperature, time on laminates were studied based orthogonal experimental method. The morphology statistically analyzed, by observing microscopic images samples cross‐section. results show that...

Summary Accurate prediction of the physical properties heterogeneous porous media based on digital models requires 3D high-resolution (HR) and large-scale images. It is, however, extremely challenging to acquire such images since current imaging technologies cannot resolve dilemma between high resolution large field view we often end up with low-resolution but a or HR small view. Moreover, available are limited always unpaired accessible Therefore, proposed hybrid unsupervised end-to-end...

In order to design and verify control algorithms for flapping wing aerial vehicles(FWAVs), calculation models of the translational force, rotational force virtual mass were established with basis on modified quasi-steady aerodynamic theory high lift mechanisms insect flight. The simulation results show that can be ignored in hovering FWAVs simple harmonic motions a cycle. effects deformation forces investigated by regarding maximum angle wingtip as reference variable. also average...

The point cloud based 3D single object tracking has drawn increasing attention. Although many breakthroughs have been achieved, we also reveal two severe issues. By extensive analysis, find the prediction manner of current approaches is non-robust, i.e., exposing a misalignment gap between score and actually localization accuracy. Another issue sparse returns will damage feature matching procedure SOT task. Based on these insights, introduce novel modules, Adaptive Refine Prediction (ARP)...

Trajectory prediction is confronted with the dilemma to capture multi-modal nature of future dynamics both diversity and accuracy. In this paper, we present a distribution discrimination (DisDis) method predict personalized motion patterns by distinguishing potential distributions. Motivated that pattern each person due his/her habit, our DisDis learns latent represent different optimize it contrastive discrimination. This encourages distributions be more discriminative. Our can integrated...