Research papers and code for "Xuan Li":
Uniformity testing and the more general identity testing are well studied problems in distributional property testing. Most previous work focuses on testing under $L_1$-distance. However, when the support is very large or even continuous, testing under $L_1$-distance may require a huge (even infinite) number of samples. Motivated by such issues, we consider the identity testing in Wasserstein distance (a.k.a. transportation distance and earthmover distance) on a metric space (discrete or continuous). In this paper, we propose the Wasserstein identity testing problem (Identity Testing in Wasserstein distance). We obtain nearly optimal worst-case sample complexity for the problem. Moreover, for a large class of probability distributions satisfying the so-called "Doubling Condition", we provide nearly instance-optimal sample complexity.

Click to Read Paper and Get Code
The infinite restricted Boltzmann machine (iRBM) is an extension of the classic RBM. It enjoys a good property of automatically deciding the size of the hidden layer according to specific training data. With sufficient training, the iRBM can achieve a competitive performance with that of the classic RBM. However, the convergence of learning the iRBM is slow, due to the fact that the iRBM is sensitive to the ordering of its hidden units, the learned filters change slowly from the left-most hidden unit to right. To break this dependency between neighboring hidden units and speed up the convergence of training, a novel training strategy is proposed. The key idea of the proposed training strategy is randomly regrouping the hidden units before each gradient descent step. Potentially, a mixing of infinite many iRBMs with different permutations of the hidden units can be achieved by this learning method, which has a similar effect of preventing the model from over-fitting as the dropout. The original iRBM is also modified to be capable of carrying out discriminative training. To evaluate the impact of our method on convergence speed of learning and the model's generalization ability, several experiments have been performed on the binarized MNIST and CalTech101 Silhouettes datasets. Experimental results indicate that the proposed training strategy can greatly accelerate learning and enhance generalization ability of iRBMs.

* Submitted to Machine Learning
Click to Read Paper and Get Code
Stochastic principal component analysis (SPCA) has become a popular dimensionality reduction strategy for large, high-dimensional datasets. We derive a simplified algorithm, called Lazy SPCA, which has reduced computational complexity and is better suited for large-scale distributed computation. We prove that SPCA and Lazy SPCA find the same approximations to the principal subspace, and that the pairwise distances between samples in the lower-dimensional space is invariant to whether SPCA is executed lazily or not. Empirical studies find downstream predictive performance to be identical for both methods, and superior to random projections, across a range of predictive models (linear regression, logistic lasso, and random forests). In our largest experiment with 4.6 million samples, Lazy SPCA reduced 43.7 hours of computation to 9.9 hours. Overall, Lazy SPCA relies exclusively on matrix multiplications, besides an operation on a small square matrix whose size depends only on the target dimensionality.

* To be published in: 2017 IEEE International Conference on Data Mining Workshops (ICDMW)
Click to Read Paper and Get Code
As the development of deep neural networks, 3D object recognition is becoming increasingly popular in computer vision community. Many multi-view based methods are proposed to improve the category recognition accuracy. These approaches mainly rely on multi-view images which are rendered with the whole circumference. In real-world applications, however, 3D objects are mostly observed from partial viewpoints in a less range. Therefore, we propose a multi-view based 3D convolutional neural network, which takes only part of contiguous multi-view images as input and can still maintain high accuracy. Moreover, our model takes these view images as a joint variable to better learn spatially correlated features using 3D convolution and 3D max-pooling layers. Experimental results on ModelNet10 and ModelNet40 datasets show that our MV-C3D technique can achieve outstanding performance with multi-view images which are captured from partial angles with less range. The results on 3D rotated real image dataset MIRO further demonstrate that MV-C3D is more adaptable in real-world scenarios. The classification accuracy can be further improved with the increasing number of view images.

* 11 pages, 11 figures
Click to Read Paper and Get Code
In the area of computer vision, deep learning has produced a variety of state-of-the-art models that rely on massive labeled data. However, collecting and annotating images from the real world has a great demand for labor and money investments and is usually too passive to build datasets with specific characteristics, such as small area of objects and high occlusion level. Under the framework of Parallel Vision, this paper presents a purposeful way to design artificial scenes and automatically generate virtual images with precise annotations. A virtual dataset named ParallelEye is built, which can be used for several computer vision tasks. Then, by training the DPM (Deformable Parts Model) and Faster R-CNN detectors, we prove that the performance of models can be significantly improved by combining ParallelEye with publicly available real-world datasets during the training phase. In addition, we investigate the potential of testing the trained models from a specific aspect using intentionally designed virtual datasets, in order to discover the flaws of trained models. From the experimental results, we conclude that our virtual dataset is viable to train and test the object detectors.

* To be published in IEEE/CAA Journal of Automatica Sinica
Click to Read Paper and Get Code
In this paper, we introduce a novel deep learning framework, termed Purine. In Purine, a deep network is expressed as a bipartite graph (bi-graph), which is composed of interconnected operators and data tensors. With the bi-graph abstraction, networks are easily solvable with event-driven task dispatcher. We then demonstrate that different parallelism schemes over GPUs and/or CPUs on single or multiple PCs can be universally implemented by graph composition. This eases researchers from coding for various parallelization schemes, and the same dispatcher can be used for solving variant graphs. Scheduled by the task dispatcher, memory transfers are fully overlapped with other computations, which greatly reduce the communication overhead and help us achieve approximate linear acceleration.

* Submitted to ICLR 2015 workshop
Click to Read Paper and Get Code
This paper proposes a forward attention method for the sequenceto- sequence acoustic modeling of speech synthesis. This method is motivated by the nature of the monotonic alignment from phone sequences to acoustic sequences. Only the alignment paths that satisfy the monotonic condition are taken into consideration at each decoder timestep. The modified attention probabilities at each timestep are computed recursively using a forward algorithm. A transition agent for forward attention is further proposed, which helps the attention mechanism to make decisions whether to move forward or stay at each decoder timestep. Experimental results show that the proposed forward attention method achieves faster convergence speed and higher stability than the baseline attention method. Besides, the method of forward attention with transition agent can also help improve the naturalness of synthetic speech and control the speed of synthetic speech effectively.

* IEEE International Conference on Acoustics, Speech and Signal Processing 2018 (ICASSP2018)
* 5 pages, 3 figures, 2 tables. Published in IEEE International Conference on Acoustics, Speech and Signal Processing 2018 (ICASSP2018)
Click to Read Paper and Get Code
Navigation services utilized by autonomous vehicles or ordinary users require the availability of detailed information about road-related objects and their geolocations, especially at road intersections. However, these road intersections are mainly represented as point elements without detailed information, or are even not available in current versions of crowdsourced mapping databases including OpenStreetMap(OSM). This study develops an approach to automatically detect road objects and place them to right location from street-level images. Our processing pipeline relies on two convolutional neural networks: the first segments the images, while the second detects and classifies the specific objects. Moreover, to locate the detected objects, we establish an attributed topological binary tree(ATBT) based on urban grammar for each image to depict the coherent relations of topologies, attributes and semantics of the road objects. Then the ATBT is further matched with map features on OSM to determine the right placed location. The proposed method has been applied to a case study in Berlin, Germany. We validate the effectiveness of our method on two object classes: traffic signs and traffic lights. Experimental results demonstrate that the proposed approach provides near-precise localization results in terms of completeness and positional accuracy. Among many potential applications, the output may be combined with other sources of data to guide autonomous vehicles

Click to Read Paper and Get Code
Depth from defocus (DfD) and stereo matching are two most studied passive depth sensing schemes. The techniques are essentially complementary: DfD can robustly handle repetitive textures that are problematic for stereo matching whereas stereo matching is insensitive to defocus blurs and can handle large depth range. In this paper, we present a unified learning-based technique to conduct hybrid DfD and stereo matching. Our input is image triplets: a stereo pair and a defocused image of one of the stereo views. We first apply depth-guided light field rendering to construct a comprehensive training dataset for such hybrid sensing setups. Next, we adopt the hourglass network architecture to separately conduct depth inference from DfD and stereo. Finally, we exploit different connection methods between the two separate networks for integrating them into a unified solution to produce high fidelity 3D disparity maps. Comprehensive experiments on real and synthetic data show that our new learning-based hybrid 3D sensing technique can significantly improve accuracy and robustness in 3D reconstruction.

Click to Read Paper and Get Code
Video image datasets are playing an essential role in design and evaluation of traffic vision algorithms. Nevertheless, a longstanding inconvenience concerning image datasets is that manually collecting and annotating large-scale diversified datasets from real scenes is time-consuming and prone to error. For that virtual datasets have begun to function as a proxy of real datasets. In this paper, we propose to construct large-scale artificial scenes for traffic vision research and generate a new virtual dataset called "ParallelEye". First of all, the street map data is used to build 3D scene model of Zhongguancun Area, Beijing. Then, the computer graphics, virtual reality, and rule modeling technologies are utilized to synthesize large-scale, realistic virtual urban traffic scenes, in which the fidelity and geography match the real world well. Furthermore, the Unity3D platform is used to render the artificial scenes and generate accurate ground-truth labels, e.g., semantic/instance segmentation, object bounding box, object tracking, optical flow, and depth. The environmental conditions in artificial scenes can be controlled completely. As a result, we present a viable implementation pipeline for constructing large-scale artificial scenes for traffic vision research. The experimental results demonstrate that this pipeline is able to generate photorealistic virtual datasets with low modeling time and high accuracy labeling.

* To be published in IEEE ITSC 2017
Click to Read Paper and Get Code
The land-use map is an important data that can reflect the use and transformation of human land, and can provide valuable reference for land-use planning. For the traditional image classification method, producing a high spatial resolution (HSR), land-use map in large-scale is a big project that requires a lot of human labor, time, and financial expenditure. The rise of the deep learning technique provides a new solution to the problems above. This paper proposes a fast and precise method that can achieve large-scale land-use classification based on deep convolutional neural network (DCNN). In this paper, we optimize the data tiling method and the structure of DCNN for the multi-channel data and the splicing edge effect, which are unique to remote sensing deep learning, and improve the accuracy of land-use classification. We apply our improved methods in the Guangdong Province of China using GF-1 images, and achieve the land-use classification accuracy of 81.52%. It takes only 13 hours to complete the work, which will take several months for human labor.

* Accepted at IEEE International Geoscience and Remote Sensing Symposium (IGARSS) 2019
Click to Read Paper and Get Code
Recent years have witnessed the great advance of deep learning in a variety of vision tasks. Many state-of-the-art deep neural networks suffer from large size and high complexity, which makes it difficult to deploy in resource-limited platforms such as mobile devices. To this end, low-precision neural networks are widely studied which quantize weights or activations into the low-bit format. Though being efficient, low-precision networks are usually hard to train and encounter severe accuracy degradation. In this paper, we propose a new training strategy through expanding low-precision networks during training and removing the expanded parts for network inference. First, we equip each low-precision convolutional layer with an ancillary full-precision convolutional layer based on a low-precision network structure, which could guide the network to good local minima. Second, a decay method is introduced to reduce the output of the added full-precision convolution gradually, which keeps the resulted topology structure the same to the original low-precision one. Experiments on SVHN, CIFAR and ILSVRC-2012 datasets prove that the proposed method can bring faster convergence and higher accuracy for low-precision neural networks.

* 10 pages, 8 figures
Click to Read Paper and Get Code
Stereo matching is the key step in estimating depth from two or more images. Recently, some tree-based non-local stereo matching methods have been proposed, which achieved state-of-the-art performance. The algorithms employed some tree structures to aggregate cost and thus improved the performance and reduced the coputation load of the stereo matching. However, the computational complexity of these tree-based algorithms is still high because they search over the entire disparity range. In addition, the extreme greediness of the minimum spanning tree (MST) causes the poor performance in large areas with similar colors but varying disparities. In this paper, we propose an efficient stereo matching method using a hierarchical disparity prediction (HDP) framework to dramatically reduce the disparity search range so as to speed up the tree-based non-local stereo methods. Our disparity prediction scheme works on a graph pyramid derived from an image whose disparity to be estimated. We utilize the disparity of a upper graph to predict a small disparity range for the lower graph. Some independent disparity trees (DT) are generated to form a disparity prediction forest (HDPF) over which the cost aggregation is made. When combined with the state-of-the-art tree-based methods, our scheme not only dramatically speeds up the original methods but also improves their performance by alleviating the second drawback of the tree-based methods. This is partially because our DTs overcome the extreme greediness of the MST. Extensive experimental results on some benchmark datasets demonstrate the effectiveness and efficiency of our framework. For example, the segment-tree based stereo matching becomes about 25.57 times faster and 2.2% more accurate over the Middlebury 2006 full-size dataset.

* 9 pages
Click to Read Paper and Get Code
Video text detection is considered as one of the most difficult tasks in document analysis due to the following two challenges: 1) the difficulties caused by video scenes, i.e., motion blur, illumination changes, and occlusion; 2) the properties of text including variants of fonts, languages, orientations, and shapes. Most existing methods attempt to enhance the performance of video text detection by cooperating with video text tracking, but treat these two tasks separately. In this work, we propose an end-to-end video text detection model with online tracking to address these two challenges. Specifically, in the detection branch, we adopt ConvLSTM to capture spatial structure information and motion memory. In the tracking branch, we convert the tracking problem to text instance association, and an appearance-geometry descriptor with memory mechanism is proposed to generate robust representation of text instances. By integrating these two branches into one trainable framework, they can promote each other and the computational cost is significantly reduced. Experiments on existing video text benchmarks including ICDAR2013 Video, Minetto and YVT demonstrate that the proposed method significantly outperforms state-of-the-art methods. Our method improves F-score by about 2 on all datasets and it can run realtime with 24.36 fps on TITAN Xp.

Click to Read Paper and Get Code
Base-detail separation is a fundamental computer vision problem consisting of modeling a smooth base layer with the coarse structures, and a detail layer containing the texture-like structures. One of the challenges of estimating the base is to preserve sharp boundaries between objects or parts to avoid halo artifacts. Many methods have been proposed to address this problem, but there is no ground-truth dataset of real images for quantitative evaluation. We proposed a procedure to construct such a dataset, and provide two datasets: Pascal Base-Detail and Fashionista Base-Detail, containing 1000 and 250 images, respectively. Our assumption is that the base is piecewise smooth and we label the appearance of each piece by a polynomial model. The pieces are objects and parts of objects, obtained from human annotations. Finally, we proposed a way to evaluate methods with our base-detail ground-truth and we compared the performances of seven state-of-the-art algorithms.

* This paper has been withdrawn by the author due to some un-proper examples
Click to Read Paper and Get Code
We study the problem of evaluating super resolution methods. Traditional evaluation methods usually judge the quality of super resolved images based on a single measure of their difference with the original high resolution images. In this paper, we proposed to use both fidelity (the difference with original images) and naturalness (human visual perception of super resolved images) for evaluation. For fidelity evaluation, a new metric is proposed to solve the bias problem of traditional evaluation. For naturalness evaluation, we let humans label preference of super resolution results using pair-wise comparison, and test the correlation between human labeling results and image quality assessment metrics' outputs. Experimental results show that our fidelity-naturalness method is better than the traditional evaluation method for super resolution methods, which could help future research on single-image super resolution.

Click to Read Paper and Get Code
We describe a novel integrated algorithm for real-time enhancement of video acquired under challenging lighting conditions. Such conditions include low lighting, haze, and high dynamic range situations. The algorithm automatically detects the dominate source of impairment, then depending on whether it is low lighting, haze or others, a corresponding pre-processing is applied to the input video, followed by the core enhancement algorithm. Temporal and spatial redundancies in the video input are utilized to facilitate real-time processing and to improve temporal and spatial consistency of the output. The proposed algorithm can be used as an independent module, or be integrated in either a video encoder or a video decoder for further optimizations.

* 10 pages, 23 figures
Click to Read Paper and Get Code
Time series forecasting is an important problem across many domains, including predictions of solar plant energy output, electricity consumption, and traffic jam situation. In this paper, we propose to tackle such forecasting problem with Transformer. Although impressed by its performance in our preliminary study, we found its two major weaknesses: (1) locality-agnostics: the point-wise dot-product self attention in canonical Transformer architecture is insensitive to local context, which can make the model prone to anomalies in time series; (2) memory bottleneck: space complexity of canonical Transformer grows quadratically with sequence length $L$, making modeling long time series infeasible. In order to solve these two issues, we first propose convolutional self attention by producing queries and keys with causal convolution so that local context can be better incorporated into attention mechanism. Then, we propose LogSparse Transformer with only $O(L(\log L)^{2})$ memory cost, improving the time series forecasting in finer granularity under constrained memory budget. Our experiments on both synthetic data and real-world datasets show that it compares favorably to the state-of-the-art.

Click to Read Paper and Get Code
In this work, we propose a novel framework named Region-Aware Network (RANet), which learns the ability of anti-confusing in case of heavy occlusion, nearby person and symmetric appearance, for human pose estimation. Specifically, the proposed method addresses three key aspects, i.e., data augmentation, feature learning and prediction fusion, respectively. First, we propose Parsing-based Data Augmentation (PDA) to generate abundant data that synthesizes confusing textures. Second, we not only propose a Feature Pyramid Stem (FPS) to learn stronger low-level features in lower stage; but also incorporate an Effective Region Extraction (ERE) module to excavate better target-specific features. Third, we introduce Cascade Voting Fusion (CVF) to explicitly exclude the inferior predictions and fuse the rest effective predictions for the final pose estimation. Extensive experimental results on two popular benchmarks, i.e. MPII and LSP, demonstrate the effectiveness of our method against the state-of-the-art competitors. Especially on easily-confusable joints, our method makes significant improvement.

Click to Read Paper and Get Code