Showing papers by "Samsung published in 2019"

A Tutorial on UAVs for Wireless Networks: Applications, Challenges, and Open Problems

Abstract: The use of flying platforms such as unmanned aerial vehicles (UAVs), popularly known as drones, is rapidly growing. In particular, with their inherent attributes such as mobility, flexibility, and adaptive altitude, UAVs admit several key potential applications in wireless systems. On the one hand, UAVs can be used as aerial base stations to enhance coverage, capacity, reliability, and energy efficiency of wireless networks. On the other hand, UAVs can operate as flying mobile terminals within a cellular network. Such cellular-connected UAVs can enable several applications ranging from real-time video streaming to item delivery. In this paper, a comprehensive tutorial on the potential benefits and applications of UAVs in wireless communications is presented. Moreover, the important challenges and the fundamental tradeoffs in UAV-enabled wireless networks are thoroughly investigated. In particular, the key UAV challenges such as 3D deployment, performance analysis, channel modeling, and energy efficiency are explored along with representative results. Then, open problems and potential research directions pertaining to UAV communications are introduced. Finally, various analytical frameworks and mathematical tools, such as optimization theory, machine learning, stochastic geometry, transport theory, and game theory are described. The use of such tools for addressing unique UAV problems is also presented. In a nutshell, this tutorial provides key guidelines on how to analyze, optimize, and design UAV-based wireless communication systems.

Highly efficient and stable InP/ZnSe/ZnS quantum dot light-emitting diodes

Abstract: Quantum dot (QD) light-emitting diodes (LEDs) are ideal for large-panel displays because of their excellent efficiency, colour purity, reliability and cost-effective fabrication1-4. Intensive efforts have produced red-, green- and blue-emitting QD-LEDs with efficiencies of 20.5 per cent4, 21.0 per cent5 and 19.8 per cent6, respectively, but it is still desirable to improve the operating stability of the devices and to replace their toxic cadmium composition with a more environmentally benign alternative. The performance of indium phosphide (InP)-based materials and devices has remained far behind those of their Cd-containing counterparts. Here we present a synthetic method of preparing a uniform InP core and a highly symmetrical core/shell QD with a quantum yield of approximately 100 per cent. In particular, we add hydrofluoric acid to etch out the oxidative InP core surface during the growth of the initial ZnSe shell and then we enable high-temperature ZnSe growth at 340 degrees Celsius. The engineered shell thickness suppresses energy transfer and Auger recombination in order to maintain high luminescence efficiency, and the initial surface ligand is replaced with a shorter one for better charge injection. The optimized InP/ZnSe/ZnS QD-LEDs showed a theoretical maximum external quantum efficiency of 21.4 per cent, a maximum brightness of 100,000 candelas per square metre and an extremely long lifetime of a million hours at 100 candelas per square metre, representing a performance comparable to that of state-of-the-art Cd-containing QD-LEDs. These as-prepared InP-based QD-LEDs could soon be usable in commercial displays.

Few-Shot Adversarial Learning of Realistic Neural Talking Head Models

Abstract: Several recent works have shown how highly realistic human head images can be obtained by training convolutional neural networks to generate them. In order to create a personalized talking head model, these works require training on a large dataset of images of a single person. However, in many practical scenarios, such personalized talking head models need to be learned from a few image views of a person, potentially even a single image. Here, we present a system with such few-shot capability. It performs lengthy meta-learning on a large dataset of videos, and after that is able to frame few- and one-shot learning of neural talking head models of previously unseen people as adversarial training problems with high capacity generators and discriminators. Crucially, the system is able to initialize the parameters of both the generator and the discriminator in a person-specific way, so that training can be based on just a few images and done quickly, despite the need to tune tens of millions of parameters. We show that such an approach is able to learn highly realistic and personalized talking head models of new people and even portrait paintings.

A Simple Baseline for Bayesian Uncertainty in Deep Learning

Abstract: We propose SWA-Gaussian (SWAG), a simple, scalable, and general purpose approach for uncertainty representation and calibration in deep learning. Stochastic Weight Averaging (SWA), which computes the first moment of stochastic gradient descent (SGD) iterates with a modified learning rate schedule, has recently been shown to improve generalization in deep learning. With SWAG, we fit a Gaussian using the SWA solution as the first moment and a low rank plus diagonal covariance also derived from the SGD iterates, forming an approximate posterior distribution over neural network weights; we then sample from this Gaussian distribution to perform Bayesian model averaging. We empirically find that SWAG approximates the shape of the true posterior, in accordance with results describing the stationary distribution of SGD iterates. Moreover, we demonstrate that SWAG performs well on a wide variety of tasks, including out of sample detection, calibration, and transfer learning, in comparison to many popular alternatives including variational inference, MC dropout, KFAC Laplace, and temperature scaling.

Emerging MPEG Standards for Point Cloud Compression

Abstract: Due to the increased popularity of augmented and virtual reality experiences, the interest in capturing the real world in multiple dimensions and in presenting it to users in an immersible fashion has never been higher. Distributing such representations enables users to freely navigate in multi-sensory 3D media experiences. Unfortunately, such representations require a large amount of data, not feasible for transmission on today’s networks. Efficient compression technologies well adopted in the content chain are in high demand and are key components to democratize augmented and virtual reality applications. Moving Picture Experts Group, as one of the main standardization groups dealing with multimedia, identified the trend and started recently the process of building an open standard for compactly representing 3D point clouds, which are the 3D equivalent of the very well-known 2D pixels. This paper introduces the main developments and technical aspects of this ongoing standardization effort.

A fully integrated reprogrammable memristor–CMOS system for efficient multiply–accumulate operations

Abstract: Memristors and memristor crossbar arrays have been widely studied for neuromorphic and other in-memory computing applications. To achieve optimal system performance, however, it is essential to integrate memristor crossbars with peripheral and control circuitry. Here, we report a fully functional, hybrid memristor chip in which a passive crossbar array is directly integrated with custom-designed circuits, including a full set of mixed-signal interface blocks and a digital processor for reprogrammable computing. The memristor crossbar array enables online learning and forward and backward vector-matrix operations, while the integrated interface and control circuitry allow mapping of different algorithms on chip. The system supports charge-domain operation to overcome the nonlinear I–V characteristics of memristor devices through pulse width modulation and custom analogue-to-digital converters. The integrated chip offers all the functions required for operational neuromorphic computing hardware. Accordingly, we demonstrate a perceptron network, sparse coding algorithm and principal component analysis with an integrated classification layer using the system. A programmable neuromorphic computing chip based on passive memristor crossbar arrays integrated with analogue and digital components and an on-chip processor enables the implementation of neuromorphic and machine learning algorithms.

Structured Knowledge Distillation for Semantic Segmentation

Abstract: In this paper, we investigate the issue of knowledge distillation for training compact semantic segmentation networks by making use of cumbersome networks. We start from the straightforward scheme, pixel-wise distillation, which applies the distillation scheme originally introduced for image classification and performs knowledge distillation for each pixel separately. We further propose to distill the structured knowledge from cumbersome networks into compact networks, which is motivated by the fact that semantic segmentation is a structured prediction problem. We study two such structured distillation schemes: (i) pair-wise distillation that distills the pairwise similarities, and (ii) holistic distillation that uses adversarial training to distill holistic knowledge. The effectiveness of our knowledge distillation approaches is demonstrated by extensive experiments on three scene parsing datasets: Cityscapes, Camvid and ADE20K.

2018 Korean Society for the Study of Obesity Guideline for the Management of Obesity in Korea.

Abstract: Obesity increases the risks of diabetes, hypertension, and cardiovascular diseases, ultimately contributing to mortality. Korean Society for the Study of Obesity (KSSO) was established to improve the management of obesity through research and education; to that end, the Committee of Clinical Practice Guidelines of KSSO reviews systemic evidence using expert panels to develop clinical guidelines. The clinical practice guidelines for obesity were revised in 2018 using National Health Insurance Service Health checkup data from 2006 to 2015. Following these guidelines, we added a category, class III obesity, which includes individuals with body mass index (BMI) ≥35 kg/m2. Agreeing with the International Federation for the Surgery of Obesity and Metabolic Disorders, Asian Pacific Chapter consensus, we determined that bariatric surgery is indicated for Korean patients with BMI ≥35 kg/m2 and for Korean patients with BMI ≥30 kg/m2 who have comorbidities. The new guidelines focus on guiding clinicians and patients to manage obesity more effectively. Our recommendations and treatment algorithms can serve as a guide for the evaluation, prevention, and management of overweight and obesity.

Intercalation-conversion hybrid cathodes enabling Li–S full-cell architectures with jointly superior gravimetric and volumetric energy densities

Abstract: A common practise in the research of Li–S batteries is to use high electrode porosity and excessive electrolytes to boost sulfur-specific capacity. Here we propose a class of dense intercalation-conversion hybrid cathodes by combining intercalation-type Mo6S8 with conversion-type sulfur to realize a Li–S full cell. The mechanically hard Mo6S8 with fast Li-ion transport ability, high electronic conductivity, active capacity contribution and high affinity for lithium polysulfides is shown to be an ideal backbone to immobilize the sulfur species and unlock their high gravimetric capacity. Cycling stability and rate capability are reported under realistic conditions of low carbon content (~10 wt%), low electrolyte/active material ratio (~1.2 µl mg−1), low cathode porosity (~55 vol%) and high mass loading (>10 mg cm−2). A pouch cell assembled based on the hybrid cathode and a 2× excess Li metal anode is able to simultaneously deliver a gravimetric energy density of 366 Wh kg−1 and a volumetric energy density of 581 Wh l−1. Despite tremendous progress in the development of LiS batteries, their performance at the full-cell level is not as competitive as state-of-the-art Li-ion batteries. Here the authors report a full-cell architecture making use of a hybrid intercalation-conversion cathode, enabling both high volumetric and gravimetric energy densities.

A wireless body area sensor network based on stretchable passive tags

Abstract: A body area sensor network (bodyNET) is a collection of networked sensors that can be used to monitor human physiological signals. For its application in next-generation personalized healthcare systems, seamless hybridization of stretchable on-skin sensors and rigid silicon readout circuits is required. Here, we report a bodyNET composed of chip-free and battery-free stretchable on-skin sensor tags that are wirelessly linked to flexible readout circuits attached to textiles. Our design offers a conformal skin-mimicking interface by removing all direct contacts between rigid components and the human body. Therefore, this design addresses the mechanical incompatibility issue between soft on-skin devices and rigid high-performance silicon electronics. Additionally, we introduce an unconventional radiofrequency identification technology where wireless sensors are deliberately detuned to increase the tolerance of strain-induced changes in electronic properties. Finally, we show that our soft bodyNET system can be used to simultaneously and continuously analyse a person’s pulse, breath and body movement. By integrating wireless stretchable on-skin sensor tags and flexible readout circuits attached to textiles using an unconventional radiofrequency identification design, a body area sensor network can be created that can continuously analyse a person’s pulse, breathing and body movement.

Learning to Quantize Deep Networks by Optimizing Quantization Intervals With Task Loss

Abstract: Reducing bit-widths of activations and weights of deep networks makes it efficient to compute and store them in memory, which is crucial in their deployments to resource-limited devices, such as mobile phones. However, decreasing bit-widths with quantization generally yields drastically degraded accuracy. To tackle this problem, we propose to learn to quantize activations and weights via a trainable quantizer that transforms and discretizes them. Specifically, we parameterize the quantization intervals and obtain their optimal values by directly minimizing the task loss of the network. This quantization-interval-learning (QIL) allows the quantized networks to maintain the accuracy of the full-precision (32-bit) networks with bit-width as low as 4-bit and minimize the accuracy degeneration with further bit-width reduction (i.e., 3 and 2-bit). Moreover, our quantizer can be trained on a heterogeneous dataset, and thus can be used to quantize pretrained networks without access to their training data. We demonstrate the effectiveness of our trainable quantizer on ImageNet dataset with various network architectures such as ResNet-18, -34 and AlexNet, on which it outperforms existing methods to achieve the state-of-the-art accuracy.

U-GAT-IT: Unsupervised Generative Attentional Networks with Adaptive Layer-Instance Normalization for Image-to-Image Translation

Abstract: We propose a novel method for unsupervised image-to-image translation, which incorporates a new attention module and a new learnable normalization function in an end-to-end manner. The attention module guides our model to focus on more important regions distinguishing between source and target domains based on the attention map obtained by the auxiliary classifier. Unlike previous attention-based method which cannot handle the geometric changes between domains, our model can translate both images requiring holistic changes and images requiring large shape changes. Moreover, our new AdaLIN (Adaptive Layer-Instance Normalization) function helps our attention-guided model to flexibly control the amount of change in shape and texture by learned parameters depending on datasets. Experimental results show the superiority of the proposed method compared to the existing state-of-the-art models with a fixed network architecture and hyper-parameters. Our code and datasets are available at this https URL or this https URL.

Vertical MoS2 Double-Layer Memristor with Electrochemical Metallization as an Atomic-Scale Synapse with Switching Thresholds Approaching 100 mV

Abstract: Atomically thin two-dimensional (2D) materials-such as transition metal dichalcogenide (TMD) monolayers and hexagonal boron nitride (hBN)-and their van der Waals layered preparations have been actively researched to build electronic devices such as field-effect transistors, junction diodes, tunneling devices, and, more recently, memristors. Two-dimensional material memristors built in lateral form, with horizontal placement of electrodes and the 2D material layers, have provided an intriguing window into the motions of ions along the atomically thin layers. On the other hand, 2D material memristors built in vertical form with top and bottom electrodes sandwiching 2D material layers may provide opportunities to explore the extreme of the memristive performance with the atomic-scale interelectrode distance. In particular, they may help push the switching voltages to a lower limit, which is an important pursuit in memristor research in general, given their roles in neuromorphic computing. In fact, recently Akinwande et al. performed a pioneering work to demonstrate a vertical memristor that sandwiches a single MoS2 monolayer between two inert Au electrodes, but it could neither attain switching voltages below 1 V nor control the switching polarity, obtaining both unipolar and bipolar switching devices. Here, we report a vertical memristor that sandwiches two MoS2 monolayers between an active Cu top electrode and an inert Au bottom electrode. Cu ions diffuse through the MoS2 double layers to form atomic-scale filaments. The atomic-scale thickness, combined with the electrochemical metallization, lowers switching voltages down to 0.1-0.2 V, on par with the state of the art. Furthermore, our memristor achieves consistent bipolar and analogue switching, and thus exhibits the synapse-like learning behavior such as the spike-timing dependent plasticity (STDP), the very first STDP demonstration among all 2D-material-based vertical memristors. The demonstrated STDP with low switching voltages is promising not only for low-power neuromorphic computing, but also from the point of view that the voltage range approaches the biological action potentials, opening up a possibility for direct interfacing with mammalian neuronal networks.

Collecting and Analyzing Multidimensional Data with Local Differential Privacy

Abstract: Local differential privacy (LDP) is a recently proposed privacy standard for collecting and analyzing data, which has been used, e.g., in the Chrome browser, iOS and macOS. In LDP, each user perturbs her information locally, and only sends the randomized version to an aggregator who performs analyses, which protects both the users and the aggregator against private information leaks. Although LDP has attracted much research attention in recent years, the majority of existing work focuses on applying LDP to complex data and/or analysis tasks. In this paper, we point out that the fundamental problem of collecting multidimensional data under LDP has not been addressed sufficiently, and there remains much room for improvement even for basic tasks such as computing the mean value over a single numeric attribute under LDP. Motivated by this, we first propose novel LDP mechanisms for collecting a numeric attribute, whose accuracy is at least no worse (and usually better) than existing solutions in terms of worst-case noise variance. Then, we extend these mechanisms to multidimensional data that can contain both numeric and categorical attributes, where our mechanisms always outperform existing solutions regarding worst-case noise variance. As a case study, we apply our solutions to build an LDP-compliant stochastic gradient descent algorithm (SGD), which powers many important machine learning tasks. Experiments using real datasets confirm the effectiveness of our methods, and their advantages over existing solutions.

Fine-Grained Analysis of Propaganda in News Article

Abstract: Propaganda aims at influencing people’s mindset with the purpose of advancing a specific agenda. Previous work has addressed propaganda detection at document level, typically labelling all articles from a propagandistic news outlet as propaganda. Such noisy gold labels inevitably affect the quality of any learning system trained on them. A further issue with most existing systems is the lack of explainability. To overcome these limitations, we propose a novel task: performing fine-grained analysis of texts by detecting all fragments that contain propaganda techniques as well as their type. In particular, we create a corpus of news articles manually annotated at fragment level with eighteen propaganda techniques and propose a suitable evaluation measure. We further design a novel multi-granularity neural network, and we show that it outperforms several strong BERT-based baselines.

Stretchable and Transparent Kirigami Conductor of Nanowire Percolation Network for Electronic Skin Applications.

Abstract: Recent research progress of relieving discomfort between electronics and human body involves serpentine designs, ultrathin films, and extraordinary properties of nanomaterials. However, these strat...

SAMSum Corpus: A Human-annotated Dialogue Dataset for Abstractive Summarization.

Abstract: This paper introduces the SAMSum Corpus, a new dataset with abstractive dialogue summaries. We investigate the challenges it poses for automated summarization by testing several models and comparing their results with those obtained on a corpus of news articles. We show that model-generated summaries of dialogues achieve higher ROUGE scores than the model-generated summaries of news -- in contrast with human evaluators' judgement. This suggests that a challenging task of abstractive dialogue summarization requires dedicated models and non-standard quality measures. To our knowledge, our study is the first attempt to introduce a high-quality chat-dialogues corpus, manually annotated with abstractive summarizations, which can be used by the research community for further studies.

Multi-scale ordering in highly stretchable polymer semiconducting films.

Abstract: Stretchable semiconducting polymers have been developed as a key component to enable skin-like wearable electronics, but their electrical performance must be improved to enable more advanced functionalities. Here, we report a solution processing approach that can achieve multi-scale ordering and alignment of conjugated polymers in stretchable semiconductors to substantially improve their charge carrier mobility. Using solution shearing with a patterned microtrench coating blade, macroscale alignment of conjugated-polymer nanostructures was achieved along the charge transport direction. In conjunction, the nanoscale spatial confinement aligns chain conformation and promotes short-range π-π ordering, substantially reducing the energetic barrier for charge carrier transport. As a result, the mobilities of stretchable conjugated-polymer films have been enhanced up to threefold and maintained under a strain up to 100%. This method may also serve as the basis for large-area manufacturing of stretchable semiconducting films, as demonstrated by the roll-to-roll coating of metre-scale films.

Learning Not to Learn: Training Deep Neural Networks With Biased Data

Abstract: We propose a novel regularization algorithm to train deep neural networks, in which data at training time is severely biased. Since a neural network efficiently learns data distribution, a network is likely to learn the bias information to categorize input data. It leads to poor performance at test time, if the bias is, in fact, irrelevant to the categorization. In this paper, we formulate a regularization loss based on mutual information between feature embedding and bias. Based on the idea of minimizing this mutual information, we propose an iterative algorithm to unlearn the bias information. We employ an additional network to predict the bias distribution and train the network adversarially against the feature embedding network. At the end of learning, the bias prediction network is not able to predict the bias not because it is poorly trained, but because the feature embedding network successfully unlearns the bias information. We also demonstrate quantitative and qualitative experimental results which show that our algorithm effectively removes the bias information from feature embedding.

Learnable Triangulation of Human Pose

Abstract: We present two novel solutions for multi-view 3D human pose estimation based on new learnable triangulation methods that combine 3D information from multiple 2D views. The first (baseline) solution is a basic differentiable algebraic triangulation with an addition of confidence weights estimated from the input images. The second, more complex, solution is based on volumetric aggregation of 2D feature maps from the 2D backbone followed by refinement via 3D convolutions that produce final 3D joint heatmaps. Crucially, both of the approaches are end-to-end differentiable, which allows us to directly optimize the target metric. We demonstrate transferability of the solutions across datasets and considerably improve the multi-view state of the art on the Human3.6M dataset.

Learnable Triangulation of Human Pose

Abstract: We present two novel solutions for multi-view 3D human pose estimation based on new learnable triangulation methods that combine 3D information from multiple 2D views. The first (baseline) solution is a basic differentiable algebraic triangulation with an addition of confidence weights estimated from the input images. The second solution is based on a novel method of volumetric aggregation from intermediate 2D backbone feature maps. The aggregated volume is then refined via 3D convolutions that produce final 3D joint heatmaps and allow modelling a human pose prior. Crucially, both approaches are end-to-end differentiable, which allows us to directly optimize the target metric. We demonstrate transferability of the solutions across datasets and considerably improve the multi-view state of the art on the Human3.6M dataset. Video demonstration, annotations and additional materials will be posted on our project page (this https URL).

Dynamic beam steering with all-dielectric electro-optic III-V multiple-quantum-well metasurfaces.

Abstract: Tunable metasurfaces enable dynamical control of the key constitutive properties of light at a subwavelength scale. To date, electrically tunable metasurfaces at near-infrared wavelengths have been realized using free carrier modulation, and switching of thermo-optical, liquid crystal and phase change media. However, the highest performance and lowest loss discrete optoelectronic modulators exploit the electro-optic effect in multiple-quantum-well heterostructures. Here, we report an all-dielectric active metasurface based on electro-optically tunable III-V multiple-quantum-wells patterned into subwavelength elements that each supports a hybrid Mie-guided mode resonance. The quantum-confined Stark effect actively modulates this volumetric hybrid resonance, and we observe a relative reflectance modulation of 270% and a phase shift from 0° to ~70°. Additionally, we demonstrate beam steering by applying an electrical bias to each element to actively change the metasurface period, an approach that can also realize tunable metalenses, active polarizers, and flat spatial light modulators.

Multi-relational Poincaré Graph Embeddings

Abstract: Hyperbolic embeddings have recently gained attention in machine learning due to their ability to represent hierarchical data more accurately and succinctly than their Euclidean analogues. However, multi-relational knowledge graphs often exhibit multiple simultaneous hierarchies, which current hyperbolic models do not capture. To address this, we propose a model that embeds multi-relational graph data in the Poincare ball model of hyperbolic space. Our Multi-Relational Poincare model (MuRP) learns relation-specific parameters to transform entity embeddings by Mobius matrix-vector multiplication and Mobius addition. Experiments on the hierarchical WN18RR knowledge graph show that our Poincare embeddings outperform their Euclidean counterpart and existing embedding methods on the link prediction task, particularly at lower dimensionality.

Ultrasoft electronics to monitor dynamically pulsing cardiomyocytes

Abstract: In biointegrated electronics, the facile control of mechanical properties such as softness and stretchability in electronic devices is necessary to minimize the perturbation of motions inherent in biological systems1–5. For in vitro studies, multielectrode-embedded dishes6–8 and other rigid devices9–12 have been widely used. Soft or flexible electronics on plastic or elastomeric substrates13–15 offer promising new advantages such as decreasing physical stress16–18 and/or applying mechanical stimuli19,20. Recently, owing to the introduction of macroporous plastic substrates with nanofibre scaffolds21,22, three-dimensional electrophysiological mapping of cardiomyocytes has been demonstrated. However, quantitatively monitoring cells that exhibit significant dynamical motions via electric probes over a long period without affecting their natural motion remains a challenge. Here, we present ultrasoft electronics with nanomeshes that monitor the field potential of human induced pluripotent stem cell-derived cardiomyocytes on a hydrogel, while enabling them to move dynamically without interference. Owing to the extraordinary softness of the nanomeshes, nanomesh-attached cardiomyocytes exhibit contraction and relaxation motions comparable to that of cardiomyocytes without attached nanomeshes. Our multilayered nanomesh devices maintain reliable operations in a liquid environment, enabling the recording of field potentials of the cardiomyocytes over a period of 96 h without significant degradation of the nanomesh devices or damage of the cardiomyocytes. Ultrasoft nanomesh electronics enable monitoring of the field potential of cardiomyocytes without interference with their natural motion.

Inkjet-printed stretchable and low voltage synaptic transistor array

Abstract: Wearable and skin electronics benefit from mechanically soft and stretchable materials to conform to curved and dynamic surfaces, thereby enabling seamless integration with the human body. However, ...

Arbitrary Shape Scene Text Detection With Adaptive Text Region Representation

Abstract: Scene text detection attracts much attention in computer vision, because it can be widely used in many applications such as real-time text translation, automatic information entry, blind person assistance, robot sensing and so on. Though many methods have been proposed for horizontal and oriented texts, detecting irregular shape texts such as curved texts is still a challenging problem. To solve the problem, we propose a robust scene text detection method with adaptive text region representation. Given an input image, a text region proposal network is first used for extracting text proposals. Then, these proposals are verified and refined with a refinement network. Here, recurrent neural network based adaptive text region representation is proposed for text region refinement, where a pair of boundary points are predicted each time step until no new points are found. In this way, text regions of arbitrary shapes are detected and represented with adaptive number of boundary points. This gives more accurate description of text regions. Experimental results on five benchmarks, namely, CTW1500, TotalText, ICDAR2013, ICDAR2015 and MSRA-TD500, show that the proposed method achieves state-of-the-art in scene text detection.

Neural Point-Based Graphics

Abstract: We present a new point-based approach for modeling the appearance of real scenes. The approach uses a raw point cloud as the geometric representation of a scene, and augments each point with a learnable neural descriptor that encodes local geometry and appearance. A deep rendering network is learned in parallel with the descriptors, so that new views of the scene can be obtained by passing the rasterizations of a point cloud from new viewpoints through this network. The input rasterizations use the learned descriptors as point pseudo-colors. We show that the proposed approach can be used for modeling complex scenes and obtaining their photorealistic views, while avoiding explicit surface estimation and meshing. In particular, compelling results are obtained for scene scanned using hand-held commodity RGB-D sensors as well as standard RGB cameras even in the presence of objects that are challenging for standard mesh-based modeling.