arXiv subjects

Kai Han

Publications and source records attributed to Kai Han.

245 records · Page 14Linked to original sources

arXiv2019

Self-calibrating Deep Photometric Stereo Networks

This paper proposes an uncalibrated photometric stereo method for non-Lambertian scenes based on deep learning. Unlike previous approaches that heavily rely on assumptions of specific reflectances and light source distributions, our method is able to determine both shape and light directions of a scene with unknown arbitrary reflectances observed under unknown varying light directions. To achieve this goal, we propose a two-stage deep learning architecture, called SDPS-Net, which can effectively take advantage of intermediate supervision, resulting in reduced learning difficulty compared to a single-stage model. Experiments on both synthetic and real datasets show that our proposed approach significantly outperforms previous uncalibrated photometric stereo methods.

cs.CV↗

arXiv2018

PS-FCN: A Flexible Learning Framework for Photometric Stereo

This paper addresses the problem of photometric stereo for non-Lambertian surfaces. Existing approaches often adopt simplified reflectance models to make the problem more tractable, but this greatly hinders their applications on real-world objects. In this paper, we propose a deep fully convolutional network, called PS-FCN, that takes an arbitrary number of images of a static object captured under different light directions with a fixed camera as input, and predicts a normal map of the object in a fast feed-forward pass. Unlike the recently proposed learning based method, PS-FCN does not require a pre-defined set of light directions during training and testing, and can handle multiple images and light directions in an order-agnostic manner. Although we train PS-FCN on synthetic data, it can generalize well on real datasets. We further show that PS-FCN can be easily extended to handle the problem of uncalibrated photometric stereo.Extensive experiments on public real datasets show that PS-FCN outperforms existing approaches in calibrated photometric stereo, and promising results are achieved in uncalibrated scenario, clearly demonstrating its effectiveness.

cs.CV↗

arXiv2018

Approximation Algorithms for Probabilistic Graphs

We study the k-median and k-center problems in probabilistic graphs. We analyze the hardness of these problems, and propose several algorithms with improved approximation ratios compared with the existing proposals.

cs.DS↗

arXiv2018

AutoEncoder Inspired Unsupervised Feature Selection

High-dimensional data in many areas such as computer vision and machine learning tasks brings in computational and analytical difficulty. Feature selection which selects a subset from observed features is a widely used approach for improving performance and effectiveness of machine learning models with high-dimensional data. In this paper, we propose a novel AutoEncoder Feature Selector (AEFS) for unsupervised feature selection which combines autoencoder regression and group lasso tasks. Compared to traditional feature selection methods, AEFS can select the most important features by excavating both linear and nonlinear information among features, which is more flexible than the conventional self-representation method for unsupervised feature selection with only linear assumptions. Experimental results on benchmark dataset show that the proposed method is superior to the state-of-the-art method.

cs.LG↗

arXiv2018

TOM-Net: Learning Transparent Object Matting from a Single Image

This paper addresses the problem of transparent object matting. Existing image matting approaches for transparent objects often require tedious capturing procedures and long processing time, which limit their practical use. In this paper, we first formulate transparent object matting as a refractive flow estimation problem. We then propose a deep learning framework, called TOM-Net, for learning the refractive flow. Our framework comprises two parts, namely a multi-scale encoder-decoder network for producing a coarse prediction, and a residual network for refinement. At test time, TOM-Net takes a single image as input, and outputs a matte (consisting of an object mask, an attenuation mask and a refractive flow field) in a fast feed-forward pass. As no off-the-shelf dataset is available for transparent object matting, we create a large-scale synthetic dataset consisting of 158K images of transparent objects rendered in front of images sampled from the Microsoft COCO dataset. We also collect a real dataset consisting of 876 samples using 14 transparent objects and 60 background images. Promising experimental results have been achieved on both synthetic and real data, which clearly demonstrate the effectiveness of our approach.

cs.CV↗

arXiv2017

Cost-Effective Seed Selection in Online Social Networks

We study the min-cost seed selection problem in online social networks, where the goal is to select a set of seed nodes with the minimum total cost such that the expected number of influenced nodes in the network exceeds a predefined threshold. We propose several algorithms that outperform the previous studies both on the theoretical approximation ratios and on the experimental performance. Under the case where the nodes have heterogeneous costs, our algorithms are the first bi- criteria approximation algorithms with polynomial running time and provable logarithmic performance bounds using a general contagion model. Under the case where the users have uniform costs, our algorithms achieve logarithmic approximation ratio and provable time complexity which is smaller than that of existing algorithms in orders of magnitude. We conduct extensive experiments using real social networks. The experimental results show that, our algorithms significantly outperform the existing algorithms both on the total cost and on the running time, and also scale well to billion-scale networks.

cs.DS↗

arXiv2017

SCNet: Learning Semantic Correspondence

This paper addresses the problem of establishing semantic correspondences between images depicting different instances of the same object or scene category. Previous approaches focus on either combining a spatial regularizer with hand-crafted features, or learning a correspondence model for appearance only. We propose instead a convolutional neural network architecture, called SCNet, for learning a geometrically plausible model for semantic correspondence. SCNet uses region proposals as matching primitives, and explicitly incorporates geometric consistency in its loss function. It is trained on image pairs obtained from the PASCAL VOC 2007 keypoint dataset, and a comparative evaluation on several standard benchmarks demonstrates that the proposed approach substantially outperforms both recent deep learning architectures and previous methods based on hand-crafted features.

cs.CV↗

arXiv2013

Truthful Scheduling Mechanisms for Powering Mobile Crowdsensing

Mobile crowdsensing leverages mobile devices (e.g., smart phones) and human mobility for pervasive information exploration and collection; it has been deemed as a promising paradigm that will revolutionize various research and application domains. Unfortunately, the practicality of mobile crowdsensing can be crippled due to the lack of incentive mechanisms that stimulate human participation. In this paper, we study incentive mechanisms for a novel Mobile Crowdsensing Scheduling (MCS) problem, where a mobile crowdsensing application owner announces a set of sensing tasks, then human users (carrying mobile devices) compete for the tasks based on their respective sensing costs and available time periods, and finally the owner schedules as well as pays the users to maximize its own sensing revenue under a certain budget. We prove that the MCS problem is NP-hard and propose polynomial-time approximation mechanisms for it. We also show that our approximation mechanisms (including both offline and online versions) achieve desirable game-theoretic properties, namely truthfulness and individual rationality, as well as O(1) performance ratios. Finally, we conduct extensive simulations to demonstrate the correctness and effectiveness of our approach.

cs.NI↗

arXiv2012

MEGCOM: Min-Energy Group COMmunication in Multi-hop Wireless Networks

Given the increasing demand from wireless applications, designing energy-efficient group communication protocols is of great importance to multi-hop wireless networks. A group communication session involves a set of member nodes, each of them needs to send a certain number of data packets to all other members. In this paper, we consider the problem of building a shared multicast tree spanning the member nodes such that the total energy consumption of a group communication session using the shared multicast tree is minimized. Since this problem was proven as NP-complete, we propose, under our Min-Energy Group COMmunication (MEGCOM) framework, three distributed approximation algorithms with provable approximation ratios. When the transmission power of each wireless node is fixed, our first two algorithms have the approximation ratios of O(ln(Δ+ 1)) and O(1), respectively, where Δis the maximum node degree in the network. When the transmission power of each wireless node is adjustable, our third algorithm again delivers a constant approximation ratio. We also use extensive simulations to verify the practical performance of our algorithms.

cs.NI↗

arXiv2012

Minimum-Energy All-to-All Multicasting in Multi-hop Wireless Networks

Designing energy-efficient all-to-all multicasting protocols is of of great importance for multi-hop wireless networks such as wireless sensor networks and wireless ad hoc networks. In an all-to-all multicast session, there exists a set of wireless destination nodes, and each destination node needs to send some data packets to all other destination nodes. We consider the problem of building a shared multicast tree spanning the destination nodes such that the total energy consumption of realizing an all-to-all multicast session using the shared multicast tree is minimized. Since building such a multicast tree has been proved to be NP-complete, we provide both centralized and distributed approximation algorithms with provable approximation ratios for it. When the transmission power of each wireless node is fixed, our centralized and distributed algorithms have the approximation ratios of $4ln(Δ+1)+7$ and 13, respectively, where $Δ$ is the maximum node degree in the network. When the transmission power of each wireless node is adjustable, both of our centralized and distributed algorithms have the constant approximation ratio of 145.

cs.NI↗

arXiv2011

Duty-Cycle-Aware Minimum-Energy Multicasting in Wireless Sensor Networks

In duty-cycled wireless sensor networks, the nodes switch between active and dormant states, and each node may determine its active/dormant schedule independently. This complicates the Minimum-Energy Multicasting (MEM) problem, which has been primarily studied in always-active wireless ad-hoc networks. In this paper, we study the duty-cycle-aware MEM problem in wireless sensor networks, and we present a formulation of the Minimum-Energy Multicast Tree Construction and Scheduling (MEMTCS) problem. We prove that the MEMTCS problem is NP-hard, and it is unlikely to have an approximation algorithm with a performance ratio of $(1-o(1))\lnΔ$, where $Δ$ is the maximum node degree in a network. We propose a polynomial-time approximation algorithm for the MEMTCS problem with a performance ratio of $\mathcal{O}(H(Δ+1))$, where $H(\cdot)$ is the harmonic number. We also provide a distributed implementation of our algorithm. Finally, we perform extensive simulations and the results demonstrate that our algorithm significantly outperform other known algorithms in terms of both the total energy cost and the transmission redundancy.

cs.NI↗