Dense Fusion Research Articles

A Multi-Scale Fusion Convolutional Neural Network Based on Attention Mechanism for the Visualization Analysis of EEG Signals Decoding.

Brain-computer interface (BCI) based on motor imagery (MI) electroencephalogram (EEG) decoding helps motor-disabled patients to communicate with external devices directly, which can achieve the purpose of human-computer interaction and assisted living. MI EEG decoding has a core problem which is extracting as many multiple types of features as possible from the multi-channel time series of EEG to understand brain activity accurately. Recently, deep learning technology has been widely used in EEG decoding. However, the variability of the simple network framework is insufficient to satisfy the complex task of EEG decoding. A multi-scale fusion convolutional neural network based on the attention mechanism (MS-AMF) is proposed in this paper. The network extracts spatio temporal multi-scale features from multi-brain regions representation signals and is supplemented by a dense fusion strategy to retain the maximum information flow. The attention mechanism we added to the network has improved the sensitivity of the network. The experimental results show that the network has a better classification effect compared with the baseline method in the BCI Competition IV-2a dataset. We conducted visualization analysis in multiple parts of the network, and the results show that the attention mechanism is also convenient for analyzing the underlying information flow of EEG decoding, which verifies the effectiveness of the MS-AMF method.

A valuable computed tomography-based new diagnostic tool for severe chest lesions in active pulmonary tuberculosis: combined application of influencing factors

The objective of the present investigation was to explore the influencing factors and value of computed tomography (CT) for diagnosing severe chest lesions in active pulmonary tuberculosis (APTB). This retrospective investigation included 463 patients diagnosed with APTB. Relevant clinical features were collected. Patients were assigned to mild/moderate group or advanced group depending on the lesion severity on chest CT, severe chest CT lesion refers to the moderately dense or less diffuse lesion that exceeds the total volume of one lung, or the dense fusion lesion greater than one third of the volume of one lung, or the lesion with cavity diameter ≥4 cm. Independent risk factors for severe lesions were determined by univariate and multivariate logistic regression analyses, and the diagnostic efficiency of the risk factors was assessed by receiver operating characteristic curve (ROC). Chest CT demonstrated that there were 285 (61.56%) cases with severe lesions; multivariate Logistic regression analysis showed dust exposure [odds ratio (OR) = 4.108, 95% confidence interval (CI) 2.416–6.986], patient classification (OR = 1.792, 95% CI 1.067–3.012), age (OR = 1.018, 95% CI 1.005–1.030), and albumin-globulin ratio (OR = 0.179, 95% CI 0.084–0.383) to be independently correlated with severe lesions on chest CT. ROC curve analysis showed the cutoff values of age, albumin-globulin ratio and combined score to be 39 years, 0.918 and −0.085, respectively. The predictive value of combined score area under the curve 0.753 (95% CI 0.708–0.798) was higher than that of any single factor. The combined score of these four factors further improved the predictive efficacy.

Deep Multi-Scale Mesh Feature Learning for Automated Labeling of Raw Dental Surfaces From 3D Intraoral Scanners.

Precisely labeling teeth on digitalized 3D dental surface models is the precondition for tooth position rearrangements in orthodontic treatment planning. However, it is a challenging task primarily due to the abnormal and varying appearance of patients' teeth. The emerging utilization of intraoral scanners (IOSs) in clinics further increases the difficulty in automated tooth labeling, as the raw surfaces acquired by IOS are typically low-quality at gingival and deep intraoral regions. In recent years, some pioneering end-to-end methods (e.g., PointNet) have been proposed in the communities of computer vision and graphics to consume directly raw surface for 3D shape segmentation. Although these methods are potentially applicable to our task, most of them fail to capture fine-grained local geometric context that is critical to the identification of small teeth with varying shapes and appearances. In this paper, we propose an end-to-end deep-learning method, called MeshSegNet, for automated tooth labeling on raw dental surfaces. Using multiple raw surface attributes as inputs, MeshSegNet integrates a series of graph-constrained learning modules along its forward path to hierarchically extract multi-scale local contextual features. Then, a dense fusion strategy is applied to combine local-to-global geometric features for the learning of higher-level features for mesh cell annotation. The predictions produced by our MeshSegNet are further post-processed by a graph-cut refinement step for final segmentation. We evaluated MeshSegNet using a real-patient dataset consisting of raw maxillary surfaces acquired by 3D IOS. Experimental results, performed 5-fold cross-validation, demonstrate that MeshSegNet significantly outperforms state-of-the-art deep learning methods for 3D shape segmentation.

Agricultural Pest Super-Resolution and Identification With Attention Enhanced Residual and Dense Fusion Generative and Adversarial Network

The growth of the most significant field crops such as rice, wheat, maize, and soybean are influenced because of various pests. And crop production is decreased due to various categories of insects. Deep learning technologies significantly increased the efficiency of identifying and controlling agricultural pests attack. However, agricultural pests images obtained are often obscure and unclear because of the sparse density of cameras deployed in the real farmland. This always makes pests difficult to recognize and monitor. Additionally, the existing classification and segmentation methods are not satisfying for the identification of low-resolution images because they are pre-trained on the clear and high-resolution datasets. Therefore, it is crucial to restore and upscale the obtained low-resolution pest images in order to improve classification accuracy and the recall rate of the instance segmentation. In this paper, we propose a generative adversarial network (GAN) with quadra-attention and residual and dense fusion mechanisms to transform low-resolution pest images. Compared with previous state-of-the-art PSNR-oriented super-resolution methods, our proposed method is more powerful in image reconstruction and achieves the state of the art performance. The experiment results show that after reconstructing with our proposed gan, the recall rate increased by 182.89% and classification accuracy also improved a lot. Besides, our proposed method could decrease the density of the camera layout in the agricultural Internet of Things (IOT) monitor systems and the cost of infrastructure, which is practical for real-world applications.

Boundary-aware RGBD Salient Object Detection with Cross-modal Feature Sampling.

Mobile devices usually mount a depth sensor to resolve ill-posed problems, like salient object detection on cluttered background. The main barrier of exploring RGBD data is to handle the information from two different modalities. To cope with this problem, in this paper, we propose a boundary-aware cross-modal fusion network for RGBD salient object detection. In particular, to enhance the fusion of color and depth features, we present a cross-modal feature sampling module to balance the contribution of the RGB and depth features based on the statistics of their channel values. In addition, in our multi-scale dense fusion network architecture, we not only incorporate edge-sensitive losses to preserve the boundary of the detected salient region, but also refine its structure by merging the estimated saliency maps of different scales. We accomplish the multi-scale saliency map merging using two alternative methods which produce refined saliency maps via per-pixel weighted combination and an encoder-decoder network. Extensive experimental evaluations demonstrate that our proposed framework can achieve the state-of-the-art performance on several public RGBD-based datasets.

Recursive Multi-Scale Image Deraining With Sub-Pixel Convolution Based Feature Fusion and Context Aggregation

Along with several other low-vision based computer vision problems, single image deraining is also taken as a challenging one due to its ill-posedness. Several algorithms based on convolutional neural networks are devised that are either too simple to provide acceptable deraining results due to under-deraining or have a complex architectures that may result into over-deraining. In this paper we propose a deraining algorithm that is capable of boosting the reconstruction/deraining quality without the problem of over or under-deraining. Along with the originally proposed network, two of its’ light-weight versions with reduced computational costs are also devised. Basically, we propose a recursively trained architecture that has two major components: a front-end module and a refinement module. The front-end module is based on dense fusion of lower label features followed by sub-pixel convolutions (pixel shuffling based convolutions). To refine and generate the enhanced deraining results further, we cascade a refinement module to the front-end module using multi-scale Context Aggregation Network (CAN) which includes feature fusion and pixel shuffling based convolutions. We present the deraining results in terms of Structural Similarity Index (SSIM) and Peak Signal to Noise Ratio (PSNR) on several benchmarks and compare with current state-of-the-art algorithms. With comprehensive experiments on both real-world and synthetic datasets and extensive ablation study, we demonstrate that our approach produces better results compared to existing methods.

Pathomorphological features of conjunctival and scleral regeneration associated with intraoperative application of Cyclosporin A

Purpose. In experiment in vivo to study the features of regeneration of the conjunctiva and sclera of rats after surgery with intraoperative application of a 0.05% Ciclosporin A.Materials and methods. Еxperimental animals (rats) (n = 48) were divided into the main group, including the subgroups a (n = 16) and b (n = 16) and the comparison group (n = 16). Performed a through cut of the conjunctiva and damage to the surface layers of the sclera one of the eyes of all animals. Further on the surgical trauma zone in the main group, the intraoperative application of the cytostatic was performed. In the subgroup a with a duration of 3 minutes, in the subgroup b – 6 minutes. In the comparison group a hemostatic sponge without a cytostatic was used intraoperatively.Results. In the comparison group postoperative period proceeds with a stereotyped dynamics of cell phase changes in damaged tissues. In the end the development of dense conjunctival-scleral fusion in the area of surgical trauma was noted. Intraoperative application of 0.05% Cyclosporine A leads to a slowing of regeneration, preventing formation of rough conjunctival-scleral scar.Conclusions. Intraoperative applications of 0.05% Cyclosporin A change the stereotyped dynamics of the inflammatory-reparative regeneration in the surgical intervention zone, inhibiting the migration of cells almost in 3 times and significantly (in 2 times) prolonging the duration of the macrophage phase. This causes a slowdown of reparative regeneration, prevents excessive scarring in the operating area.

基于特征密集计算与融合算法的教师课堂行为分析

基于特征密集计算与融合算法的教师课堂行为分析

Deep Spatiotemporal Relation Learning With 3D Multi-Level Dense Fusion for Video Action Recognition

Two-stream Convolutional Neural Network has shown a remarkable performance for video action recognition. Many recent works mainly focus on the fusion of appearance and motion information to obtain a robust spatiotemporal representation for action video. However, most of these networks are based on 2D convolution architecture and apply spatiotemporal fusion at the top layers of the network, which lacks the capability to take full advantage of the potential spatiotemporal relation in multiple levels of the network as well as capture the temporal dynamic in low-level details. In this paper, we propose a novel convolutional fusion network based on a two-stream network, called 3D Multi-Level Dense Fusion (MLDF-3D) for the deep spatiotemporal relation learning. There are mainly three merits of the proposed network: (i) rather than performing fusion only at the last convolution layer or softmax layer, the MLDF-3D performs spatiotemporal fusion at multiple levels of network, which can fully explore the potential relation of features extracted from two-stream network; (ii) we introduce dense connection in MLDF-3D to make fusion features at multiple levels directly connected; and (iii) we develop a sequence segment framework for the long-range temporal structure modeling. Our experimental results show that the MLDF-3D can effectively learn the correlation of two-stream features to obtain robust action representation, and achieve the state-of-the-art performance on HMDB51 and UCF101.

Hydro-dissection in a post hysterectomy female with dense labial fusion: a case report

Hydro-dissection in a post hysterectomy female with dense labial fusion: a case report

Real-time dense map fusion for stereo SLAM

SUMMARYA robot should be able to estimate an accurate and dense 3D model of its environment (a map), along with its pose relative to it, all of it in real time, in order to be able to navigate autonomously without collisions.As the robot moves from its starting position and the estimated map grows, the computational and memory footprint of a dense 3D map increases and might exceed the robot capabilities in a short time. However, a global map is still needed to maintain its consistency and plan for distant goals, possibly out of the robot field of view.In this work, we address such problem by proposing a real-time stereo mapping pipeline, feasible for standard CPUs, which is locally dense and globally sparse and accurate. Our algorithm is based on a graph relating poses and salient visual points, in order to maintain a long-term accuracy with a small cost. Within such framework, we propose an efficient dense fusion of several stereo depths in the locality of the current robot pose.We evaluate the performance and the accuracy of our algorithm in the public datasets of Tsukuba and KITTI, and demonstrate that it outperforms single-view stereo depth. We release the code as open-source, in order to facilitate the system use and comparisons.

A compact new incoherent Thomson scattering diagnostic for low-temperature plasma studies

Incoherent Thomson scattering (ITS) has a long history of application for the determination of electron density and temperature in dense fusion plasmas, and in recent years, has been increasingly extended to studies in low-temperature plasma environments. In this work, the design and preliminary implementation of a new, sensitive and uniquely compact ITS platform known as Thomson scattering experiments for low temperature ion sources are described. Measurements have been performed on a hollow cathode plasma source, providing access to electron densities as low as 1016 m−3 and electron temperatures of a few eV and below. This achievement has been made possible by the implementation of a narrow volume Bragg grating notch filter for the attenuation of stray light, a feature which guarantees compactness and reduced transmission losses in comparison to standard ITS platforms.

Excitation of half-integer up-shifted decay channel and quasi-mode in plasma edge for high power electron Bernstein wave heating scenario

Electron Bernstein waves (EBW) consist of promising tools in driving localized off-axis current needed for sustained operation as well as effective selective heating scenarios in advanced over dense fusion plasmas like spherical tori and stellarators by applying high power radio frequency waves within the range of Megawatts. Here some serious non-linear effects like parametric decay modes are highly expect-able which have been extensively studied theoretically and experimentally. In general, the decay of an EBW depends on the ratio of the incident frequency and electron cyclotron frequency. At ratios less than two, parametric decay leads to a lower hybrid wave (or an ion Bernstein wave) and EBWs at a lower frequency. For ratios more than two, the daughter waves constitute either an electron cyclotron quasi-mode and another EBW or an ion wave and EBW. However, in contrast with these decay patterns, the excitation of an unusual up-shifted frequency decay channel for the ratio less than two is demonstrated in this study which is totally different as to its generation and persistence. It is shown that this mode varies from the conventional parametric decay channels which necessarily satisfy the matching conditions in frequency and wave-vector. Moreover, the excitation of some less-known local non-propagating quasi-modes (virtual modes) through weak-turbulence theory and their contributions to energy leakage from conversion process leading the reduction in conversion efficiency is assessed.

Micro-scale fusion in dense relativistic nanowire array plasmas

Nuclear fusion is regularly created in spherical plasma compressions driven by multi-kilojoule pulses from the world’s largest lasers. Here we demonstrate a dense fusion environment created by irradiating arrays of deuterated nanostructures with joule-level pulses from a compact ultrafast laser. The irradiation of ordered deuterated polyethylene nanowires arrays with femtosecond pulses of relativistic intensity creates ultra-high energy density plasmas in which deuterons (D) are accelerated up to MeV energies, efficiently driving D–D fusion reactions and ultrafast neutron bursts. We measure up to 2 × 106 fusion neutrons per joule, an increase of about 500 times with respect to flat solid targets, a record yield for joule-level lasers. Moreover, in accordance with simulation predictions, we observe a rapid increase in neutron yield with laser pulse energy. The results will impact nuclear science and high energy density research and can lead to bright ultrafast quasi-monoenergetic neutron point sources for imaging and materials studies.

The Photogrammetric Reconstruction Task Solved by Bundle Adjustment Directly with Pixel Bundles

At present the reconstruction task of photogrammetry is performed mostly in several steps: one feature-based bundle solution followed by many dense image matchings and fusion of intermediate results. To increase the performance of reconstruction to a maximum a different definition is proposed, starting from the approach of strict digital image inversion and the correspondence condition of image grey values with the unknown function of object surface intensity (or brightness). This leads again to a bundle solution (similar to the classical solution), however, directly with the image grey values as measurements so that each pixel represents a ray and each image a high-density bundle of rays (pixel bundle). All images of a block are forced to generate optimally one function of object surface intensity together with the object surface and the orientation of the images (WROBEL, Proceedings SPIE 804, pp 325–334. In German language: Bildmessung und Luftbildwesen 1987). The diverse tasks of modern life need different methods of 3D reconstructions in diverging performance. Here, we concentrate on a high-fidelity approach of both the functional and stochastic model of image evaluation. The fundamental derivations are presented, main properties (e.g. its statistical potential) are discussed and highlighted in comparison to deficiencies of existing approaches. Topographical reconstructions demonstrate the validity of the results with respect to the surface and the standard deviations of its heights.

Effect of low-intensity pulsed ultrasound on posterolateral lumbar fusion of rabbit.

To observe the effect of low-intensity pulsed ultrasound in the posterolateral lumbar fusion of rabbit. A total of 48 New Zealand white rabbits were randomly divided into the observation group and control group, including the autogenous iliac observation group, artificial bone observation group, artificial bone control group and autogenous iliac control group according to the different bone grafting. The posterolateral lumbar fusion of rabbits in each group was analyzed and compared. After 4 weeks of treatment using the low-intensity pulsed ultrasound, the fusion in the bone grafting area of observation group was good. There was the relatively dense fusion area between the right transverse process and artificial bone. The left transverse process had been completely fused, with the clear bone trabecula through the fusion area. There was no significant fusion sign in the control group. According to the fusion comparison between two groups, the fusion rate of the observation group was 83.3% and it was significantly higher than the one of the control group (37.5%). Results of the comparison were statistically significant (P<0.05). The number of chondrocytes and the increase in the relative gray-scale in the fusion area of the iliac observation group were significantly higher than ones of the iliac control group (P<0.05). The number of chondrocytes and the increase in the relative gray-scale in the fusion area of the artificial bone observation group were significantly higher than ones of the artificial bone control group (P<0.05). The expression of IL-1, IL-6 and IL-17 4 weeks after the treatment of the observation group was significantly lower than the one of the control group, with the statistical significance (P<0.05). It indicated that the treatment of low-intensity pulsed ultrasound could reduce the expression level of inflammatory factors. The low-intensity pulsed ultrasound can significantly increase the bone grafting fusion rate of the rabbit's posterolateral lumbar fusion. The possible mechanism is that it promotes the lumbar posterolateral endochondral ossification and reduces the inflammatory reaction.

The Effect of Energy Leakage Probability on Burn Propagation in an Optically Thick Fusion Plasma

In an optically thick plasma, the mean free path of bremsstrahlung photons is smaller than the plasma radius, and radiation can be treated as a photon gas in thermal equilibrium. In these conditions, the black body radiation spectrum exceeds the number of hot photons, and reabsorption processes such as inverse bremsstrahlung radiation and inverse Compton scattering become important. It has been shown that a dense fusion plasma like the one being used in ICF method is initially optically thick. When the fuel pellet is burning, the temperature of its electrons rises (approximately greater than 90 KeV), and the pellet becomes rapidly optically thin. In this paper, we have shown that the energy leakage probability makes electron temperature remain low (approximately smaller than 55 KeV), and as a result the fuel pellet remains optically thick during burning.

Evidence of anomalous resistivity for hot electron propagation through a dense fusion core in fast ignition experiments

Anomalous resistivity for hot electrons passing through a dense core plasma is studied for fast ignition laser fusion. The hot electrons generated via the ultra-intense laser pulse and guiding cone interactions are measured after they pass through a dense plasma with a density of 50–100 g cm−3 in a radius of 15–25 μm. When significant neutron enhancements are achieved by the ultra-intense laser pulse injection, the energy reduction of fast electrons is observed. Also, a reduction in the number of electrons with energy up to 15 MeV can be seen. We offer a new physical mechanism for the stopping of electrons, involving electron magnetohydrodynamic shock formation in the inhomogeneous plasma density region. The dissipation in the shock region can explain electron stopping with energies of the order of 15 MeV.

Plans for Warm Dense Matter and IFE Target Experiments on NDCX-II

The Heavy Ion Fusion Science Virtual National Laboratory (HIFS-VNL) is currently developing design concepts for NDCX-II, the second phase of the Neutralized Drift Compression Experiment, which will use ion beams to explore Warm Dense Matter (WDM) and Inertial Fusion Energy (IFE) target hydrodynamics. The ion induction accelerator will consist of a new short pulse injector and induction cells from the decommissioned Advanced Test Accelerator (ATA) at Lawrence Livermore National Laboratory (LLNL). To fit within an existing building and to meet the energy and temporal requirements of various target experiments, an aggressive beam compression and acceleration schedule is planned. WDM physics and ion-driven direct drive hydrodynamics will initially be explored with 30 nC of lithium ions in experiments involving ion deposition, ablation, acceleration and stability of planar targets. Other ion sources which may deliver higher charge per bunch will be explored. A test stand has been built at Lawrence Berkeley National Laboratory (LBNL) to test refurbished ATA induction cells and pulsed power hardware for voltage holding and ability to produce various compression and acceleration waveforms. Another test stand is being used to develop and characterize lithium-doped aluminosilicate ion sources. The first experiments will include heating metallic targets to 10,000 K and hydrodynamics studies with cryogenic hydrogen targets.

Neutrino plasma coupling in dense astrophysical plasmas

There is considerable interest in the propagation dynamics of intense neutrino beams in a background dispersive medium such as dense plasmas, particularly in the search for a mechanism to explain the dynamics of type II supernovae. Neutrino interactions with matter are usually considered as single particle interactions. All the single particle mechanisms describing the dynamical properties of neutrinos in matter are analogous with the processes involving single electron interactions with a medium such as Compton scattering, Cerenkov radiation, etc. However, it is well known that beams of electrons moving through a plasma give rise to a new class of processes known as collective interactions, such as two stream instabilities, which result in either the absorption or generation of plasma waves. Employing the relativistic kinetic equations for neutrinos interacting with dense plasmas via the weak force, we explore collective plasma streaming instabilities driven by neutrino beams. We examine the anomalous transfer between neutrinos and dense plasma via excitation of electron plasma waves. The nonlinear coupling between an intense neutrino beam and a plasma reveals the presence of two regimes, a hydrodynamic regime and a kinetic regime. The latter is responsible for Landau damping or growth of electron plasma waves. In dense fusion stellar plasmas neutrino Landau damping can play a significant role as an additional stellar plasma cooling process. Another interesting result is an asymmetry in the momentum balance imported by the neutrinos to the core of the exploding star due to symmetry breaking by the collapsed star's magnetic fields. This results in a directed velocity of the resulting neutron star or pulsar, explaining the so called ‘birth’ velocity.