Wei-Ming Wang

Wei-Ming Wang
Shanghai Jiao Tong University | SJTU · Department of Mechanical Engineering (ME)

Professor

About

65
Publications
9,278
Reads
How we measure 'reads'
A 'read' is counted each time someone views a publication summary (such as the title, abstract, and list of authors), clicks on a figure, or views or downloads the full-text. Learn more
475
Citations
Introduction
My research interests include robotic grasping and manipulation, physical human-robot interaction.
Skills and Expertise

Publications

Publications (65)
Preprint
Full-text available
In this paper, we tackle the problem of category-level 9D pose estimation in the wild, given a single RGB-D frame. Using supervised data of real-world 9D poses is tedious and erroneous, and also fails to generalize to unseen scenarios. Besides, category-level pose estimation requires a method to be able to generalize to unseen objects at test time,...
Article
Point cloud analysis without pose priors is very challenging in real applications, as the orientations of point clouds are often unknown. In this paper, we propose a brand new point-set learning framework PRIN, namely, Point-wise Rotation Invariant Network, focusing on rotation invariant feature extraction in point clouds analysis. We construct sph...
Preprint
Pixel-level 2D object semantic understanding is an important topic in computer vision and could help machine deeply understand objects (e.g. functionality and affordance) in our daily life. However, most previous methods directly train on correspondences in 2D images, which is end-to-end but loses plenty of information in 3D spaces. In this paper,...
Article
Robotic picking of diverse range of novel objects is a great challenge in dense clutter, in which objects are stacked together tightly. However, collecting large-scale dataset with dense grasp labels is extremely time-consuming, and there is huge gap between synthetic color and depth images with real images. In this paper, we explore suction based...
Article
Pixel-level 2D object semantic understanding is an important topic in computer vision and could help machine deeply understand objects (e.g. functionality and affordance) in our daily life. However, most previous methods directly train on correspondences in 2D images, which is end-to-end but loses plenty of information in 3D spaces. In this paper,...
Preprint
Full-text available
Point cloud analysis without pose priors is very challenging in real applications, as the orientations of point clouds are often unknown. In this paper, we propose a brand new point-set learning framework PRIN, namely, Point-wise Rotation Invariant Network, focusing on rotation invariant feature extraction in point clouds analysis. We construct sph...
Article
Full-text available
Real-time vision-based robotic grasping is challenging in clutter. In such scene, the target object should be perceived accurately, where it may be occluded and misrecognized by many distractors including irrelevant objects and the robotic arm. In addition, the limited field of view (FOV) of camera makes it prone for objects to get out of the camer...
Article
Recently, plenty of deep learning methods have been proposed to handle point clouds. Almost all of them input the entire point cloud and ignore the information redundancy lying in point clouds. This paper addresses this problem by extracting the Reeb graph from point clouds, which is a much more informative and compact representation of point cloud...
Preprint
Full-text available
Keypoint detection is an essential component for the object registration and alignment. However, previous works mainly focused on how to register keypoints under arbitrary rigid transformations. Differently, in this work, we reckon keypoints under an information compression scheme to represent the whole object. Based on this, we propose UKPGAN, an...
Preprint
Full-text available
3D object detection has attracted much attention thanks to the advances in sensors and deep learning methods for point clouds. Current state-of-the-art methods like VoteNet regress direct offset towards object centers and box orientations with an additional Multi-Layer-Perceptron network. Both their offset and orientation predictions are not accura...
Chapter
Semantic understanding of 3D objects is crucial in many applications such as object manipulation. However, it is hard to give a universal definition of point-level semantics that everyone would agree on. We observe that people have a consensus on semantic correspondences between two areas from different objects, but are less certain about the exact...
Article
This paper proposes a methodology to generate online smooth joint trajectories of robots based on an improved sinusoidal jerk model. The multi-segment trajectory model is designed to allow a more sufficient exploitation of the actuation capability, thereby shortening the execution time while ensuring continuity up to the jerk level. Afterwards, the...
Article
Full-text available
Robotic grasping of diverse range of novel objects is a great challenge in dense clutter, which is also critical to many applications. However, current methods are vulnerable to perception uncertainty for dense stacked objects, resulting in limited accuracy of multi-parameter grasp prediction. In this paper, we propose a two-stage grasp detection p...
Article
Full-text available
The development of the hardware design of robot body and external sensors enables the improvement of multiple perception and control methods for efficient interactions with humans and unstructured environments. An important application area for integrating humans and robots with such advanced interactions is physical Human-Robot Interaction (pHRI)....
Preprint
Full-text available
Visual semantic correspondence is an important topic in computer vision and could help machine understand objects in our daily life. However, most previous methods directly train on correspondences in 2D images, which is end-to-end but loses plenty of information in 3D spaces. In this paper, we propose a new method on predicting semantic correspond...
Article
Full-text available
Point cloud analysis without pose priors is very challenging in real applications, as the orientations of point clouds are often unknown. In this paper, we propose a brand new point-set learning framework PRIN, namely, Pointwise Rotation-Invariant Network, focusing on rotation-invariant feature extraction in point clouds analysis. We construct sphe...
Preprint
Detecting 3D objects keypoints is of great interest to the areas of both graphics and computer vision. There have been several 2D and 3D keypoint datasets aiming to address this problem in a data-driven way. These datasets, however, either lack scalability or bring ambiguity to the definition of keypoints. Therefore, we present KeypointNet: the fir...
Article
Full-text available
Deep learning-based visuomotor control for manipulation are studied recently, which uses a neural network to learn the mapping from images to robotic actions directly. Most previous methods assume there is only one target object in the image. When multiple target objects are present, they are normally regarded as multi-task problems. One-hot vector...
Preprint
Full-text available
Fine-grained semantic understanding of 3D objects is crucial in many applications such as object manipulation. However, it is hard to give a universal definition of point-level semantics that everyone would agree on. We observe that people are pretty sure about semantic correspondences between two areas from different objects, but less certain abou...
Article
Full-text available
Fast abnormal event detection meets the growing demand to process an enormous number of surveillance videos. Based on the inherent redundancy of video structures, we propose an efficient sparse combination learning framework with both batch and online solvers. It achieves decent performance in the detection phase without compromising result quality...
Preprint
Full-text available
In this paper we study grasp problem in dense cluster, a challenging task in warehouse logistics scenario. By introducing a two-step robust suction affordance detection method, we focus on using vacuum suction pad to clear up a box filled with seen and unseen objects. Two CNN based neural networks are proposed. A Fast Region Estimation Network (FRE...
Preprint
Full-text available
Robotic grasp detection is a fundamental capability for intelligent manipulation in unstructured environments. Previous work mainly employed visual and tactile fusion to achieve stable grasp, while, the whole process depending heavily on regrasping, which wastes much time to regulate and evaluate. We propose a novel way to improve robotic grasping:...
Preprint
Full-text available
This paper focuses on robotic picking tasks in cluttered scenario. Because of the diversity of poses, types of stack and complicated background in bin picking situation, it is much difficult to recognize and estimate their pose before grasping them. Here, this paper combines Resnet with U-net structure, a special framework of Convolution Neural Net...
Preprint
Full-text available
Deep reinforcement learning (DRL) has gained a lot of attention in recent years, and has been proven to be able to play Atari games and Go at or above human levels. However, those games are assumed to have a small fixed number of actions and could be trained with a simple CNN network. In this paper, we study a special class of Asian popular card ga...
Article
Full-text available
Deep neural network-based end-to-end visuomotor control for robotic manipulation is becoming a hot issue of robotics field recently. One-hot vector is often used for multi-task situation in this framework. However, it is inflexible using one-hot vector to describe multiple tasks and transmit intentions of humans. This paper proposes a framework by...
Article
Confocal laser scanning microscopy (CLSM) has emerged as one of the most advanced fluorescence cell imaging techniques in the field of biomedicine. However, fluorescence cell imaging is limited by spatial blur and additive white noise induced by the excitation light. In this paper, a spatially adaptive high-order total variation (SA-HOTV) model for...
Preprint
Full-text available
In recent years, point clouds have earned quite some research interests by the development of depth sensors. Due to different layouts of objects, orientation of point clouds is often unknown in real applications. In this paper, we propose a new point sets learning framework named Pointwise Rotation-Invariant Network (PRIN), focusing on the rotation...
Article
Full-text available
Planning efficient trajectories is an essential task in most automated robotic applications. The execution time and smoothness are usually important considerations for economic and safety reasons. A novel method to generate a trigonometric frequency central pattern generator trajectory is presented in this paper for cyclic point-to-point tasks of i...
Article
Full-text available
We propose a deep learning approach for directly estimating relative atmospheric visibility from outdoor photos, which does not rely on weather images or data that require expensive sensing or custom capture. Our data-driven approach capitalizes on a large collection of internet images to learn rich scene and visibility varieties. The relative CNN-...
Article
Full-text available
Electromyogram (EMG) signal decoding is the essential part of myoelectric control. However, traditional machine learning methods lack the capability of learning and expressing the information contained in EMG signals, and the robustness of the myoelectric control system is not sufficient for real life applications. In this article, a novel model ba...
Poster
Full-text available
This poster considers the problem of picking objects in cluster. This requires the robot to reliably detect the picking point for the known or unseen objects under the environment with occlusion, disorder and a variety of objects. We present a novel pipeline to detect picking point based on deep convolutional neural network (CNN). At last, we demon...
Conference Paper
Functional modeling has been widely researched in the past decades and is considered to be effective in assisting concept design. However, construction of substantial design repository with abundant knowledge maintenance is still a problem. in this paper, we propose a model and approach for retrieve functional knowledge from existing products. Firs...
Article
Electrical evoked potentials (EEPs) time series prediction is a novel topic concentrating on reducing the cost of the visual prostheses research. Support vector machine (SVM), a superior neural network algorithm, is a powerful tool for time series forecasting but is insensitive to multivariate analysis. Meanwhile, similarity measurement (SM), a key...
Article
In customer-driven design, reusing the design experiences of solving previous problems is a potential methodology, and the case retrieval (CR) process is a major step process, in which similarity measurement (SM) among cases is its core. However, performing the CR model with high retrieval accuracy and low computational complexity for the fuzzy, va...
Article
Case retrieval (CR) is critically an important part of case-based design. However few studies attempt to research CR for customer-driven design and analyze the effect of other production factors besides similarity computation. This paper proposes a new CR method for customer-driven design, and requirement-weighting analysis. Fuzzy set theory are in...
Article
Functionality represents a blueprint of a product and plays a crucial role in problem-solving such as design. This article discusses the model representation from the angle of functional ontology by function deployment. We construct a framework of functional ontology which decomposes the function and contains a library of vocabulary to comprehensiv...
Article
Full-text available
In current practice, the design is still based on the sequential design methodology, which makes the process or manufacture information not considered at the preliminary design stage. In addition, the quality of design and the time to perform them are largely dependent on the experience of the engineers. Case-based design is an intelligent method w...
Article
Uncertainties exist in every aspect of a collaborative multidisciplinary design process. These uncertainties will have a great influence on design negotiations between various disciplines and may force designers to make conservative decisions. In this paper, a novel collaborative robust optimization (CRO) method based on constraints network under u...
Article
Process conflict and forepart conflict were presented in the automotive configuration. Focusing on the process conflict, this paper took mathematical model, customized requirements and knowledge as configuration items, and then constraint model was built. Conflict effect weight was proposed, and constraint loosen based conflict strategy was used in...
Article
A collaborative parameter robust optimization based on generalized dynamic constraints network (GDCN) was presented. The uncertainty of parameter in the concurrent and collaborative design was analyzed, and the GDCN was presented to manage the uncertainty. The frame of collaborative parameter robust design based on GDCN was developed. The model of...
Article
Full-text available
Through defining the conception of transforming-table and operation-function, aiming at various kinds of mapping types, this paper proposed a mapping method for BOM (bill of material) multi-view. The method records all the mapping parameters of all kinds of BOM through transforming-table, and fulfills the consistency mapping process through the mai...
Article
One of the main issues of the reverse engineering is the duplication of an existing physical part whose geometric information is partially or completely unavailable in measured form. In some industrial applications, physical parts are duplicated using three-axis CNC machines and ball end-mills. Many researches studied the problem of direct tool pat...
Article
Domain knowledge is very significant to process design in metal plastic forming area. The paper presents a knowledge representation, fusion and management method for metal-forming process. Firstly, a web-based ontology description language is used to describe the knowledge and a knowledge fusion and knowledge reusing framework is proposed in metal-...
Article
This paper discusses rule-based reasoning parameter consistency management system based on knowledge in concurrent and corroborative product design using a combination of both mathematical methods and knowledge-based Engineering (KBE) techniques. First, the framework of the parameter consistency management system is developed. Second, a data-mining...
Article
Artificial intelligence and knowledge based engineering technologies are widely used in die and mould design. In this paper an intelligent deep-drawing process design system for complex circular shells is developed, which build the product representation based on shape element and the knowledge base by machine learning technology. It helps the engi...
Article
In view of uncertainty of parameters in concurrent and collaborative design, the constraints network is extended to the generalized dynamic constraints network (GDCN) to cope with the uncertainty trouble. First, the model of GDCN including both the domain level and knowledge level constraints is established. Second, the effective method to represen...
Conference Paper
This paper describes an uncertainty management method using generalized dynamic constraint networks (GDCN). Uncertainty of parameter in the concurrent and collaborative design is analyzed, and GDCN is presented to manage the uncertainty. Firstly, the model of GDCN including domain level constraints and knowledge level constraints is established. Se...
Conference Paper
This paper presents a parameter coordination approach based on constraints network to support multidisciplinary collaborative design. Firstly, from the point of view of collaborative design, the model of constraints network for parameter coordination is studied. In this model, interval boxes are adopted to describe the uncertainty of design paramet...
Article
An effective modeling method of domain level constraints in the constraint network for concurrent engineering (CE) was developed. The domain level constraints were analyzed and the framework of modeling of domain level constraints based on simulation and approximate technology was given. An intelligent response surface methodology (IRSM) was propos...
Article
The concurrent and collaborative design of the complicated product based on generalized dynamic constraints network (GDCN) was presented, and the mathematical model of the GDCN was also presented. Then, the modeling method of GDCN was analyzed, and the effective modeling method of the domain constraints based on simulation and Meta modeling was pre...
Conference Paper
This paper discusses the development of a knowledge-based constraints network system for concurrent product design using a combination of both mathematic methods and knowledge-based engineering (KBE) techniques. Firstly, the model of knowledge-based constraints network system (KCNS) is developed. Secondly, a data-mining algorithm named fuzzy-rough...
Conference Paper
This research work aims to develop a GDCN (generalized dynamic constraint network) that enables designers to consider at the early stages of the design process all activities associated with product's life cycle. This research article discusses the development of effective modeling method for domain level constraints using a combination of both num...
Article
A computational fluid dynamics (CFD) simulation for analyzing fluid flow patterns in a plasma spray gun is presented in this study. It is coupled with a heat transfer simulation of the plasma spray gun. Based on CFD and heat transfer theory, the numerical model of the nozzle in the plasma spray gun is developed, and the coupled simulation of the fl...
Article
The cooling performance at the key regions of mold during the low pressure die casting (LPDC) of magnesium wheel was investigated. Through analyzing the shrinkage defects generated under the conditions of various cooling modes, it was found the cooling pipe system set in the side mold alone is a valid way to enhance the cooling capacity at the rim/...
Conference Paper
This research article demonstrates the use of a constraints network for modeling the knowledge which is necessary for concurrent product design. A Knowledge-based Constraints Network System (KCNS) has been developed to maintain design consistency and to support the selection of appropriate design parameter intervals. A data-mining algorithm named f...
Article
Based on CFD and heat transfer theory, a model of the nozzle in a plasma sprayer is developed. The flow and heat transfer of the cooling water are simulated with coupling. It is found that the wall near heat source and the gasket in the back of the nozzle are the most deficient cooling parts. Then different parameters of cooling water are analyzed,...
Conference Paper
Parameter and tolerance design is two important phases in product design. The design process of a complex product should consider synthetically manufacturing and assembly process. In this paper, firstly, agent based on constraints network modeling approach for assembly and manufacturing process is presented to model concurrent product development....
Article
Based on Computational Fluid Dynamics (CFD) and Semi-Implicit Method for Pressure-Linked-Equations (SIMPLE) method, a model of plasma sprayed gun was developed and the flow field of cooling water was simulated. The local reflow and turbulent, which will lead to static-water region, are found in the front of the cooling water cavity. The structure o...

Network

Cited By