Wei Dong

Bio: Wei Dong is an academic researcher from Zhejiang University. The author has contributed to research in topics: Wireless sensor network & Network packet. The author has an hindex of 32, co-authored 177 publications receiving 34953 citations. Previous affiliations of Wei Dong include Princeton University & University of Texas at Austin.

ImageNet: A large-scale hierarchical image database

Abstract: The explosion of image data on the Internet has the potential to foster more sophisticated and robust models and algorithms to index, retrieve, organize and interact with images and multimedia data. But exactly how such data can be harnessed and organized remains a critical problem. We introduce here a new database called “ImageNet”, a large-scale ontology of images built upon the backbone of the WordNet structure. ImageNet aims to populate the majority of the 80,000 synsets of WordNet with an average of 500-1000 clean and full resolution images. This will result in tens of millions of annotated images organized by the semantic hierarchy of WordNet. This paper offers a detailed analysis of ImageNet in its current state: 12 subtrees with 5247 synsets and 3.2 million images in total. We show that ImageNet is much larger in scale and diversity and much more accurate than the current image datasets. Constructing such a large-scale database is a challenging task. We describe the data collection scheme with Amazon Mechanical Turk. Lastly, we illustrate the usefulness of ImageNet through three simple applications in object recognition, image classification and automatic object clustering. We hope that the scale, accuracy, diversity and hierarchical structure of ImageNet can offer unparalleled opportunities to researchers in the computer vision community and beyond.

Efficient k-nearest neighbor graph construction for generic similarity measures

Abstract: K-Nearest Neighbor Graph (K-NNG) construction is an important operation with many web related applications, including collaborative filtering, similarity search, and many others in data mining and machine learning. Existing methods for K-NNG construction either do not scale, or are specific to certain similarity measures. We present NN-Descent, a simple yet efficient algorithm for approximate K-NNG construction with arbitrary similarity measures. Our method is based on local search, has minimal space overhead and does not rely on any shared global index. Hence, it is especially suitable for large-scale applications where data structures need to be distributed over the network. We have shown with a variety of datasets and similarity measures that the proposed method typically converges to above 90% recall with each point comparing only to several percent of the whole dataset on average.

Does wireless sensor network scale? A measurement study on GreenOrbs

Abstract: In spite of the remarkable efforts the community put to build the sensor systems, an essential question still remains unclear at the system level, motivating us to explore the answer from a point of real-world deployment view. Does the wireless sensor network really scale? We present findings from a large scale operating sensor network system, GreenOrbs, with up to 330 nodes deployed in the forest. We instrument such an operating network throughout the protocol stack and present observations across layers in the network. Based on our findings from the system measurement, we propose and make initial efforts to validate three conjectures that give potential guidelines for future designs of large scale sensor networks. (1) A small portion of nodes bottlenecks the entire network, and most of the existing network indicators may not accurately capture them. (2) The network dynamics mainly come from the inherent concurrency of network operations instead of environment changes. (3) The environment, although the dynamics are not as significant as we assumed, has an unpredictable impact on the sensor network. We suggest that an event-based routing structure can be trained optimal and thus better adapt to the wild environment when building a large scale sensor network.

Secure friend discovery in mobile social networks

Abstract: Mobile social networks extend social networks in the cyberspace into the real world by allowing mobile users to discover and interact with existing and potential friends who happen to be in their physical vicinity. Despite their promise to enable many exciting applications, serious security and privacy concerns have hindered wide adoption of these networks. To address these concerns, in this paper we develop novel techniques and protocols to compute social proximity between two users to discover potential friends, which is an essential task for mobile social networks.We make three major contributions. First, we identify a range of potential attacks against friend discovery by analyzing real traces. Second, we develop a novel solution for secure proximity estimation, which allows users to identify potential friends by computing social proximity in a privacy-preserving manner. A distinctive feature of our solution is that it provides both privacy and verifiability, which are frequently at odds in secure multiparty computation. Third, we demonstrate the feasibility and effectiveness of our approaches using real implementation on smartphones and show it is efficient in terms of both computation time and power consumption.

Rapid: A Multimodal and Device-free Approach Using Noise Estimation for Robust Person Identification

Abstract: Device-free human sensing is a key technology to support many applications such as indoor navigation and activity recognition. By exploiting WiFi signals reflected by human body, there have been many WiFi-based device-free human sensing applications. Among these applications, person identification is a fundamental technology to enable user-specific services. In this paper, we present Rapid, a system that can perform robust person identification in a device-free and low-cost manner, using fine-grained channel information (i.e., CSI) of WiFi and acoustic information from footstep sound. In order to achieve high accuracy in real-life scenarios with both system and environment noise, we perform noise estimation and include two different confidence values to quantify the impact of noise to both CSI and acoustic measurements. Based on an accurate gait analysis, we then adaptively fuse CSI and acoustic measurements to achieve robust person identification. We implement low-cost Rapid nodes and evaluate our system using experiments at multiple locations with a total of 1800 gait instances from 20 volunteers, and the results show that Rapid identifies a subject with an average accuracy of 92% to 82% from a group of 2 to 6 subjects, respectively.

ImageNet Classification with Deep Convolutional Neural Networks

Abstract: We trained a large, deep convolutional neural network to classify the 1.2 million high-resolution images in the ImageNet LSVRC-2010 contest into the 1000 different classes. On the test data, we achieved top-1 and top-5 error rates of 37.5% and 17.0% which is considerably better than the previous state-of-the-art. The neural network, which has 60 million parameters and 650,000 neurons, consists of five convolutional layers, some of which are followed by max-pooling layers, and three fully-connected layers with a final 1000-way softmax. To make training faster, we used non-saturating neurons and a very efficient GPU implementation of the convolution operation. To reduce overriding in the fully-connected layers we employed a recently-developed regularization method called "dropout" that proved to be very effective. We also entered a variant of this model in the ILSVRC-2012 competition and achieved a winning top-5 test error rate of 15.3%, compared to 26.2% achieved by the second-best entry.

Very Deep Convolutional Networks for Large-Scale Image Recognition

Abstract: In this work we investigate the effect of the convolutional network depth on its accuracy in the large-scale image recognition setting. Our main contribution is a thorough evaluation of networks of increasing depth using an architecture with very small (3x3) convolution filters, which shows that a significant improvement on the prior-art configurations can be achieved by pushing the depth to 16-19 weight layers. These findings were the basis of our ImageNet Challenge 2014 submission, where our team secured the first and the second places in the localisation and classification tracks respectively. We also show that our representations generalise well to other datasets, where they achieve state-of-the-art results. We have made our two best-performing ConvNet models publicly available to facilitate further research on the use of deep visual representations in computer vision.

Very Deep Convolutional Networks for Large-Scale Image Recognition

Abstract: In this work we investigate the effect of the convolutional network depth on its accuracy in the large-scale image recognition setting. Our main contribution is a thorough evaluation of networks of increasing depth using an architecture with very small (3x3) convolution filters, which shows that a significant improvement on the prior-art configurations can be achieved by pushing the depth to 16-19 weight layers. These findings were the basis of our ImageNet Challenge 2014 submission, where our team secured the first and the second places in the localisation and classification tracks respectively. We also show that our representations generalise well to other datasets, where they achieve state-of-the-art results. We have made our two best-performing ConvNet models publicly available to facilitate further research on the use of deep visual representations in computer vision.

Going deeper with convolutions

Abstract: We propose a deep convolutional neural network architecture codenamed Inception that achieves the new state of the art for classification and detection in the ImageNet Large-Scale Visual Recognition Challenge 2014 (ILSVRC14). The main hallmark of this architecture is the improved utilization of the computing resources inside the network. By a carefully crafted design, we increased the depth and width of the network while keeping the computational budget constant. To optimize quality, the architectural decisions were based on the Hebbian principle and the intuition of multi-scale processing. One particular incarnation used in our submission for ILSVRC14 is called GoogLeNet, a 22 layers deep network, the quality of which is assessed in the context of classification and detection.

Deep Learning

Abstract: Deep learning is a form of machine learning that enables computers to learn from experience and understand the world in terms of a hierarchy of concepts. Because the computer gathers knowledge from experience, there is no need for a human computer operator to formally specify all the knowledge that the computer needs. The hierarchy of concepts allows the computer to learn complicated concepts by building them out of simpler ones; a graph of these hierarchies would be many layers deep. This book introduces a broad range of topics in deep learning. The text offers mathematical and conceptual background, covering relevant concepts in linear algebra, probability theory and information theory, numerical computation, and machine learning. It describes deep learning techniques used by practitioners in industry, including deep feedforward networks, regularization, optimization algorithms, convolutional networks, sequence modeling, and practical methodology; and it surveys such applications as natural language processing, speech recognition, computer vision, online recommendation systems, bioinformatics, and videogames. Finally, the book offers research perspectives, covering such theoretical topics as linear factor models, autoencoders, representation learning, structured probabilistic models, Monte Carlo methods, the partition function, approximate inference, and deep generative models. Deep Learning can be used by undergraduate or graduate students planning careers in either industry or research, and by software engineers who want to begin using deep learning in their products or platforms. A website offers supplementary material for both readers and instructors.