ALBERT

Description: Multiple-license plate recognition is gaining popularity in the Intelligent Transport System (ITS) applications for security monitoring and surveillance. Advancements in acquisition devices have increased the availability of high definition (HD) images, which can capture images of multiple vehicles. Since license plate (LP) occupies a relatively small portion of an image, therefore, detection of LP in an image is considered a challenging task. Moreover, the overall performance deteriorates when the aforementioned factor combines with varying illumination conditions, such as night, dusk, and rainy. As it is difficult to locate a small object in an entire image, this paper proposes a two-step approach for plate localization in challenging conditions. In the first step, the Faster-Region-based Convolutional Neural Network algorithm (Faster R-CNN) is used to detect all the vehicles in an image, which results in scaled information to locate plates. In the second step, morphological operations are employed to reduce non-plate regions. Meanwhile, geometric properties are used to localize plates in the HSI color space. This approach increases accuracy and reduces processing time. For character recognition, the look-up table (LUT) classifier using adaptive boosting with modified census transform (MCT) as a feature extractor is used. Both proposed plate detection and character recognition methods have significantly outperformed conventional approaches in terms of precision and recall for multiple plate recognition.

Description: Depth is essential information for autonomous robotics applications that need environmental depth values. The depth could be acquired by finding the matching pixels between stereo image pairs. Depth information is an inference from a matching cost volume that is composed of the distances between the possible pixel points on the pre-aligned horizontal axis of stereo images. Most approaches use matching costs to identify matches between stereo images and obtain depth information. Recently, researchers have been using convolutional neural network-based solutions to handle this matching problem. In this paper, a novel method has been proposed for the refinement of matching costs by using recurrent neural networks. Our motivation is to enhance the depth values obtained from matching costs. For this purpose, to attain an enhanced disparity map by utilizing the sequential information of matching costs in the horizontal space, recurrent neural networks are used. Exploiting this sequential information, we aimed to determine the position of the correct matching point by using recurrent neural networks, as in the case of speech processing problems. We used existing stereo algorithms to obtain the initial matching costs and then improved the results by utilizing recurrent neural networks. The results are evaluated on the KITTI 2012 and KITTI 2015 datasets. The results show that the matching cost three-pixel error is decreased by an average of 14.5% in both datasets.

Description: Perfect image compositing can harmonize the appearance between the foreground and background effectively so that the composite result looks seamless and natural. However, the traditional convolutional neural network (CNN)-based methods often fail to yield highly realistic composite results due to overdependence on scene parsing while ignoring the coherence of semantic and structural between foreground and background. In this paper, we propose a framework to solve this problem by training a stacked generative adversarial network with attention guidance, which can efficiently create a high-resolution, realistic-looking composite. To this end, we develop a diverse adversarial loss in addition to perceptual and guidance loss to train the proposed generative network. Moreover, we construct a multi-scenario dataset for high-resolution image compositing, which contains high-quality images with different styles and object masks. Experiments on the synthesized and real images demonstrate the efficiency and effectiveness of our network in producing seamless, natural, and realistic results. Ablation studies show that our proposed network can improve the visual performance of composite results compared with the application of existing methods.

Description: Recent attention to facial alignment and landmark detection methods, particularly with application of deep convolutional neural networks, have yielded notable improvements. Neither these neural-network nor more traditional methods, though, have been tested directly regarding performance differences due to camera-lens focal length nor camera viewing angle of subjects systematically across the viewing hemisphere. This work uses photo-realistic, synthesized facial images with varying parameters and corresponding ground-truth landmarks to enable comparison of alignment and landmark detection techniques relative to general performance, performance across focal length, and performance across viewing angle. Recently published high-performing methods along with traditional techniques are compared in regards to these aspects.

Description: Touchless fingerprint recognition represents a rapidly growing field of research which has been studied for more than a decade. Through a touchless acquisition process, many issues of touch-based systems are circumvented, e.g., the presence of latent fingerprints or distortions caused by pressing fingers on a sensor surface. However, touchless fingerprint recognition systems reveal new challenges. In particular, a reliable detection and focusing of a presented finger as well as an appropriate preprocessing of the acquired finger image represent the most crucial tasks. Also, further issues, e.g., interoperability between touchless and touch-based fingerprints or presentation attack detection, are currently investigated by different research groups. Many works have been proposed so far to put touchless fingerprint recognition into practice. Published approaches range from self identification scenarios with commodity devices, e.g., smartphones, to high performance on-the-move deployments paving the way for new fingerprint recognition application scenarios.This work summarizes the state-of-the-art in the field of touchless 2D fingerprint recognition at each stage of the recognition process. Additionally, technical considerations and trade-offs of the presented methods are discussed along with open issues and challenges. An overview of available research resources completes the work.

Description: As High Efficiency Video Coding (HEVC) is a worldwide popular video coding standard, the steganography of HEVC videos has gained more and more attention. Prediction unit (PU) is one of the most important innovative modules of HEVC; thus, PU partition mode-based steganography is becoming a novel branch of HEVC steganography. However, the embedding capacity of this kind of steganography is limited by the types of PU partition modes. To solve the problem, modified exploiting modification direction (EMD)-coded PU partition mode-based steganography is proposed in this paper, which can hide a secret digit in a (2n + x − 1)-ary notational system in a pair of PU partition modes and thus enlarging the capacity. Furthermore, two mapping patterns for PU partition modes are analyzed, and the one that performs the better is selected as the final mapping pattern. Firstly, 8 × 8- and 16 × 16-sized PU partition modes are recorded according to the optimal mapping pattern in the video encoding process. Then, PU partition modes are modified by using the proposed method to satisfy the requirement of secret information. Finally, the stego video can be obtained by re-encoding the video with the modified PU partition modes. Experimental results show that the embedding capacity can be significantly enlarged, and compared with the state-of-the-art work, the proposed method has much larger capacity while keeping high visual quality.

Description: Due to the prevalence of social media service, effective and efficient online image retrieval is in urgent need to satisfy diversified requirements of Web users. Previous studies are mainly focusing on bridging the semantic gap by well-established content modeling with semantic information and social tagging information, but they are not flexible in aggregating the diversified expectations of the online users. In this paper, we present OSIR, a solution framework to facilitate the diversified preference styles in online social media image searching by textual query inputs. First, we propose an efficient Online Multiple Kernel Ranking (OMKR) model which is constructed on multiple query dimensions and complimentary feature channels, and trained by minimizing the triplet loss on hard negative samples. By optimizing the ranking performance with multi-dimensional queries, the semantic consistency between the image ranking and textual query input is directly maximized without relying on the intermediate semantic annotation procedure. Second, we construct random walk-based preference modeling by domain-specific similarity calculation on heterogeneous social attributes. By re-ranking the rank output of OMKR based on each preference ranking model, we obtain a set of ranking lists encoding different potential aspects of user preference. Last, we propose an effective and efficient position-sensitive rank aggregation approach to aggregate multiple ranking results based on the user preference specification. Extensive experiment on two social media datasets demonstrates the advantages of our approach in both retrieval performance and user experience.

Description: Chinese named entity recognition (CNER) in the judicial domain is an important and fundamental task in the analysis of judgment documents. However, only a few researches have been devoted to this task so far. For Chinese named entity recognition in judgment documents, we propose the use a bidirectional long-short-term memory (BiLSTM) model, which uses character vectors and sentence vectors trained by distributed memory model of paragraph vectors (PV-DM). The output of BiLSTM is used by conditional random field (CRF) to tag the input sequence. We also improved the Viterbi algorithm to increase the efficiency of the model by cutting the path with the lowest score. At last, a novel dataset with manual annotations is constructed. The experimental results on our corpus show that the proposed method is effective not only in reducing the computational time, but also in improving the effectiveness of named entity recognition in the judicial domain.

Description: Artificial intelligence has been widely studied on solving intelligent surveillance analysis and security problems in recent years. Although many multimedia security approaches have been proposed by using deep learning network model, there are still some challenges on their performances which deserve in-depth research. On the one hand, high computational complexity of current deep learning methods makes it hard to be applied to real-time scenario. On the other hand, it is difficult to obtain the specific features of a video by fine-tuning the network online with the object state of the first frame, which fails to capture rich appearance variations of the object. To solve above two issues, in this paper, an effective object tracking method with learning attention is proposed to achieve the object localization and reduce the training time in adversarial learning framework. First, a prediction network is designed to track the object in video sequences. The object positions of the first ten frames are employed to fine-tune prediction network, which can fully mine a specific features of an object. Second, the prediction network is integrated into the generative adversarial network framework, which randomly generates masks to capture object appearance variations via adaptively dropout input features. Third, we present a spatial attention mechanism to improve the tracking performance. The proposed network can identify the mask that maintains the most robust features of the objects over a long temporal span. Extensive experiments on two large-scale benchmarks demonstrate that the proposed algorithm performs favorably against state-of-the-art methods.

Description: In the image inpainting method based on sparse representation, the adaptability of over-complete dictionary has a great influence on the result of image restoration. If the over-complete dictionary cannot effectively reflect the differences between different local features, it may result in the loss of texture details, resulting in blurred or over-smooth phenomenon in restored images. In view of these problems, we propose an image restoration method based on sparse representation using feature classification learning. Firstly, we perform singular value decomposition on the local gradient vector. According to the relationship between the main orientation and the secondary orientation, we classify all the local patches into three categories: smooth patch, edge patch and texture patch. Secondly, we use K-Singular Value Decomposition method to learn over-complete dictionaries that adapt to different features. Finally, we use Orthogonal Matching Pursuit method to calculate the sparse coding of target patches with different local features on their corresponding over-complete dictionaries, and use the over-complete dictionary and corresponding sparse coding to restore the damaged pixels. A series of experiments on various restoration tasks show the superior performance of the proposed method.