Dense Connections Research Articles

A very lightweight image super-resolution network

Recently, ConvNeXt and blueprint separable convolution (BSConv) constructed from standard ConvNet modules have demonstrated competitive performance in advanced computer vision tasks. This paper proposes an efficient model (BCRN) based on BSConv and the ConvNeXt residual structure for single image super-resolution, which achieves superior performance with very low parametric numbers. Specifically, the residual block (BCB) of the BCRN utilizes the ConvNeXt residual structure and BSConv to significantly reduce the number of parameters. Within the residual block, enhanced spatial attention and contrast-aware channel attention modules are simultaneously introduced to prioritize valuable features within the network. Multiple residual blocks are then stacked to form the backbone network, with Dense connections utilized between them to enhance feature utilization. Our model boasts extremely low parameters compared to other state-of-the-art lightweight models, while experimental results on benchmark datasets demonstrate its excellent performance. The code will be available at https://github.com/kptx666/BCRN.

Lightweight network for millimeter-level concrete crack detection with dense feature connection and dual attention

Development of lightweight deep learning crack detection method is essential for the future deployment of mobile device-based structure inspection. The primary challenge involves the analysis and extraction of features from narrow cracks, typically 3–6 pixels wide, which are often obscured by noise such as water stains and shadows. The lightweight model should also maintain high accuracy while ensuring low computational complexity and a minimal number of parameters. To this end, this paper proposes YOLO v5-DE (Dense Feature Enhancement Connection, Efficient and Fast Convolution), a lightweight network based on the YOLO v5 architecture tailored to address these challenges, and constructs crack datasets captured at different heights to investigate the impact of different shooting distances on network performance. The network utilizes efficient convolutions and dense feature connections, with strategic reuse of filtered features from shallow layers, to significantly enhance the model's fine-grained feature information and gradient flow. The experimental results demonstrate that YOLO v5-DE achieves a detection accuracy of 96% for cracks in concrete structures. Compared to the improved YOLO v5 with EfficientViT as the backbone network, YOLO v5-DE achieves 4.7% increase on accuracy while requiring fewer computational resources, with only 1.4 million parameters and 3.6 Giga Floating point Operations Per Second (GFLOPS). Additionally, YOLO v5-DE reduces the inference time to 3.38 ms and increases the frame rate to 295.8 FPS. Moreover, the proposed lightweight network exhibits better detection performance when facing complex backgrounds and real-world environments.

The fusion of lock-in phase images for the damage investigation in composites materials

Lock-in thermography (LIT) is a type of active thermography capable of detecting and evaluating the subsurface defects in composite materials. The phase or amplitude difference between the defect and sound regions can quantitatively determine the size and depth of defects. One limitation of LIT is that the optimal identification of defects located at specific depths is dispersed within phase images acquired at varying excitation frequencies. If the depth of the defects within the specimen is unknown, it is difficult to determine the excitation frequencies to achieve optimal contrast for all defects and display them in a single image. To address this challenge, a multi-frequency fused method based on lock-in thermography is proposed to improve the quality of defect detection. Initially, the optimal thermal wave excitation frequencies and the number of detection times are determined based on the theoretical solution. Subsequently, phase images at various excitation frequencies are extracted and enhanced using a specified scheme. Finally, we develop an unsupervised encoder-decoder network that combines dense connections and residual modules to fuse these phase images into a single image containing defects at various depths. An experiment is conducted to detect carbon fiber reinforced polymer (CFRP) laminate containing defects with different depths and sizes. The results demonstrate that the proposed method can broaden the detection range of defect depths and improve the quality of the inspected images, providing superior performance compared to traditional sequential image processing methods.

Deep‐learning‐based downscaling of precipitation in the middle reaches of the Yellow River using residual‐based CNNs

AbstractThe middle reaches of the Yellow River (MRYR), located in northern China, are the transition zone between semi‐arid and semi‐humid climates. As one of the climate‐sensitive regions in China, MRYR has a fragile ecological environment and serious soil loss, which leads to geological disasters such as landslides, collapses, and mudslides caused by extreme precipitation. However, scarceness of high‐resolution precipitation data over MRYR limits assessment of the environmental impacts caused by climate change, especially for extreme precipitation. In this article, we design a Residual‐in‐Residual Dense Block based Network (RRDBNet) model for the statistical downscaling of precipitation in MRYR, and compare the proposed RRDBNet with a generalized linear regression model (GLM) and two popular deep‐learning‐based models. The multi‐level residuals and dense connectivity strategies introduced in RRDBNet help it to learn more abstract features and complex nonlinear relationships among climate variables to improve downscaling performance. The results show that the proposed RRDBNet has good performance in precipitation simulations, which can reproduce the spatial–temporal characteristics of high‐resolution precipitation well. RRDBNet reduces the root‐mean‐squared error (RMSE) by 19% and improves the Pearson correlation coefficient (CC) by 6% relative to GLM for climatology mean precipitation. Especially, RRDBNet has substantial improvements in extreme precipitation compared with other models. It reduces RMSE by 58% (79%) and improves CC by 38% (145%) relative to GLM for R95P (R99P), where R95P and R99P represent extreme precipitation and very extreme precipitation, respectively. For the probability density function of daily precipitation, it is further demonstrated that RRDBNet performs better as regards extreme precipitation frequency. Our results suggest that statistical downscaling based on RRDBNet may be an effective tool for historical and future climate simulations from global climate models.

Self-supervised monocular depth estimation with self-distillation and dense skip connection

Monocular depth estimation (MDE) is crucial in a wide range of applications, including robotics, autonomous driving and virtual reality. Self-supervised monocular depth estimation has emerged as a promising MDE approach without requiring hard-to-obtain depth labels during training, and multi-scale photometric loss is widely used for self-supervised monocular depth estimation as the self-supervised signal. However, multi-photometric loss is a weak training signal and might disturb the good intermediate features representation. In this paper, we propose a successive depth map self-distillation(SDM-SD) loss, which combines with the single-scale photometric loss to replace the multi-scale photometric loss. Moreover, considering that multi-stage feature representations are essential for dense prediction tasks such as depth estimation, we also propose a dense skip connection, which can efficiently fuse the intermediate features of the encoder and fully utilize them in each stage of the decoder in our encoder–decoder architecture. By applying successive depth map self-distillation loss and dense skip connection, our proposed method can achieve state-of-the-art performance on the KITTI benchmark, and exhibit the best generalization ability on the challenging indoor dataset NYUv2 dataset.

White matter fibre density in the brain's inhibitory control network is associated with falling in low activity older adults.

Recent research has indicated that the relationship between age-related cognitive decline and falling may be mediated by the individual's capacity to quickly cancel or inhibit a motor response. This longitudinal investigation demonstrates that higher white matter fibre density in the motor inhibition network paired with low physical activity was associated with falling in elderly participants. We measured the density of white matter fibre tracts connecting key nodes in the inhibitory control network in a large sample (n = 414) of older adults. We modelled their self-reported frequency of falling over a 4-year period with white matter fibre density in pathways corresponding to the direct and hyperdirect cortical-subcortical loops implicated in the inhibitory control network. Only connectivity between right inferior frontal gyrus and right subthalamic nucleus was associated with falling as measured cross-sectionally. The connectivity was not, however, predictive of future falling when measured 2 and 4 years later. Higher white matter fibre density was associated with falling, but only in combination with low levels of physical activity. No such relationship existed for selected control brain regions that are not implicated in the inhibitory control network. Albeit statistically robust, the direction of this effect was counterintuitive (more dense connectivity associated with falling) and warrants further longitudinal investigation into whether white matter fibre density changes over time in a manner correlated with falling, and mediated by physical activity.

LDCNet: A Lightweight Multiscale Convolutional Neural Network Using Local Dense Connectivity for Image Recognition

LDCNet: A Lightweight Multiscale Convolutional Neural Network Using Local Dense Connectivity for Image Recognition

Brain tumor segmentation based on the U-NET+⁣+ network with efficientnet encoder.

Brain tumor is a highly destructive, aggressive, and fatal disease. The presence of brain tumors can disrupt the brain's ability to control body movements, consciousness, sensations, thoughts, speech, and memory. Brain tumors are often accompanied by symptoms like epilepsy, headaches, and sensory loss, leading to varying degrees of cognitive impairment in affected patients. The study goal is to develop an effective method to detect and segment brain tumor with high accurancy. This paper proposes a novel U-Net+⁣+ network using EfficientNet as the encoder to segment brain tumors based on MRI images. We adjust the original U-Net+⁣+ model by removing the dense skip connections between sub-networks to simplify computational complexity and improve model efficiency, while the connections of feature maps at the same resolution level are retained to bridge the semantic gap. The proposed segmentation model is trained and tested on Kaggle's LGG brain tumor dataset, which obtains a satisfying performance with a Dice coefficient of 0.9180. This paper conducts research on brain tumor segmentation, using the U-Net+⁣+ network with EfficientNet as an encoder to segment brain tumors based on MRI images. We adjust the original U-Net+⁣+ model to simplify calculations and maintains rich semantic spatial features at the same time. Multiple loss functions are compared in this study and their effectiveness are discussed. The experimental results shows the model achieves a high segmention result with Dice coefficient of 0.9180.

Modelling a dense network connectivity for panoptic tooth segmentation using learning approaches

By enabling precise diagnoses and assisting in creating successful treatment regimens, teeth segmentation plays a crucial role in dentistry. The segmentation of teeth has been the primary emphasis of previous approaches, but they frequently ignore the context of the oral tissue as a whole. According to this study, background semantic and foreground instance-based segmentation results are combined in a panoptic-segmentation-based approach. This work provides a tailored architecture for segmenting using a Super Adaptive Dense Convolutional Neural Network (sa−DCNN) with a panoptic quality (PQ) loss function. The PQ loss function speeds up learning and enables the model to predict masks and the classes that go with them instantly. In our suggested design, sa−DCNN enables connectivity between the memory path and the sa−DCNN pixel route. A stacked network block integrating multi-scale properties from several decoding resolutions is also included. The network block's primary duties involve integrating various mechanisms. The output heads analyze the memory path and pixel route data to predict the mask classes used to create the final mask. The proposed model outperforms the most recent panoptic segmentation on the UFBA-UESC Dental Image dataset, notably in performance and resilience. Our study represents a significant advancement in teeth segmentation and advances our knowledge of oral anatomy.

Intrinsic Dynamics and Neural Implementation of a Hypothalamic Line Attractor Encoding an Internal Behavioral State.

Line attractors are emergent population dynamics hypothesized to encode continuous variables such as head direction and internal states. In mammals, direct evidence of neural implementation of a line attractor has been hindered by the challenge of targeting perturbations to specific neurons within contributing ensembles. Estrogen receptor type 1 (Esr1)-expressing neurons in the ventrolateral subdivision of the ventromedial hypothalamus (VMHvl) show line attractor dynamics in male mice during fighting. We hypothesized that these dynamics may encode continuous variation in the intensity of an internal aggressive state. Here, we report that these neurons also show line attractor dynamics in head-fixed mice observing aggression. We exploit this finding to identify and perturb line attractor-contributing neurons using 2-photon calcium imaging and holographic optogenetic perturbations. On-manifold perturbations demonstrate that integration and persistent activity are intrinsic properties of these neurons which drive the system along the line attractor, while transient off-manifold perturbations reveal rapid relaxation back into the attractor. Furthermore, stimulation and imaging reveal selective functional connectivity among attractor-contributing neurons. Intriguingly, individual differences among mice in line attractor stability were correlated with the degree of functional connectivity among contributing neurons. Mechanistic modelling indicates that dense subnetwork connectivity and slow neurotransmission are required to explain our empirical findings. Our work bridges circuit and manifold paradigms, shedding light on the intrinsic and operational dynamics of a behaviorally relevant mammalian line attractor.

Frequency-Separated Attention Network for Image Super-Resolution

The use of deep convolutional neural networks has significantly improved the performance of super-resolution. Employing deeper networks to enhance the non-linear mapping capability from low-resolution (LR) to high-resolution (HR) images has inadvertently weakened the information flow and disrupted long-term memory. Moreover, overly deep networks are challenging to train, thus failing to exhibit the expressive capability commensurate with their depth. High-frequency and low-frequency features in images play different roles in image super-resolution. Networks based on CNNs, which should focus more on high-frequency features, treat these two types of features equally. This results in redundant computations when processing low-frequency features and causes complex and detailed parts of the reconstructed images to appear as smooth as the background. To maintain long-term memory and focus more on the restoration of image details in networks with strong representational capabilities, we propose the Frequency-Separated Attention Network (FSANet), where dense connections ensure the full utilization of multi-level features. In the Feature Extraction Module (FEM), the use of the Res ASPP Module expands the network’s receptive field without increasing its depth. To differentiate between high-frequency and low-frequency features within the network, we introduce the Feature-Separated Attention Block (FSAB). Furthermore, to enhance the quality of the restored images using heuristic features, we incorporate attention mechanisms into the Low-Frequency Attention Block (LFAB) and the High-Frequency Attention Block (HFAB) for processing low-frequency and high-frequency features, respectively. The proposed network outperforms the current state-of-the-art methods in tests on benchmark datasets.

Double reuses based residual network

Deep residual network (ResNet) had shown remarkable performance in image recognition tasks, due to its shorter connections between layers close to the input and those close to the output. And densely connected convolutional network (DenseNet) had further improved recognition performance by dense feature reuses. To improve the performance of residual units in ResNet and feature reuses in DenseNet, we propose a simple and effective convolutional network architecture, named double reuses based residual network (DRRNet). DRRNet improves the residual unit of ResNet, where the feature reuse connections are added to combine all feature maps from the convolutional layers to produce the residual, and uses a residual reuse path outside units to reuse all residuals as the final feature maps for classification. Residual learning used in DRRNet can alleviate the vanishing-gradient problem. The double reuses including inner-unit feature reuses and outer-unit residual reuses effectively decrease computational cost as compared with dense connections in DenseNet, and further strengthen the forward feature propagation. DRRNet is evaluated on three object recognition benchmark datasets and an object detection dataset. In comparison with the state-of-the-art, DRRNet achieves a good balance between classification accuracy and computational cost, and optimal detection performance.

CompleteDT: Point cloud completion with information-perception transformers

In this work, we propose a novel point cloud completion network, called CompleteDT. To fully capture the 3D geometric structure of point clouds, we introduce an Information-Perception Transformer (IPT) that can simultaneously capture local features and global geometric relations. CompleteDT comprises a Feature Encoder, Query Generator, and Query Decoder. Feature Encoder extracts local features from multi-resolution point clouds to capture intricate geometrical structures. Query Generator uses the proposed IPT, utilizing the Point Local Attention (PLA) and Point Global Attention (PGA) modules, to learn local features and global correlations, and generate query features that represent predicted point clouds. The PLA captures local information within local points by adaptively measuring weights of neighboring points, while PGA adapts multi-head self-attention by transforming it into a layer-by-layer form where each head learns global features in a high-dimensional space of different dimensions. By dense connections, the module allows for direct information exchange between each head and facilitates the capture of long global correlations. By combining the strengths of both PLA and PGA, the IPT can fully leverage local and global features to facilitate CompleteDT to complete point clouds. Lastly, the query features undergo refining to generate a complete point cloud through the Query Decoder. Our experimental results demonstrate that CompleteDT outperforms current state-of-the-art methods, effectively learning from incomplete inputs and predicting complete outputs.

The meso-connectomes of mouse, marmoset, and macaque: network organization and the emergence of higher cognition.

The recent publications of the inter-areal connectomes for mouse, marmoset, and macaque cortex have allowed deeper comparisons across rodent vs. primate cortical organization. In general, these show that the mouse has very widespread, "all-to-all" inter-areal connectivity (i.e. a "highly dense" connectome in a graph theoretical framework), while primates have a more modular organization. In this review, we highlight the relevance of these differences to function, including the example of primary visual cortex (V1) which, in the mouse, is interconnected with all other areas, therefore including other primary sensory and frontal areas. We argue that this dense inter-areal connectivity benefits multimodal associations, at the cost of reduced functional segregation. Conversely, primates have expanded cortices with a modular connectivity structure, where V1 is almost exclusively interconnected with other visual cortices, themselves organized in relatively segregated streams, and hierarchically higher cortical areas such as prefrontal cortex provide top-down regulation for specifying precise information for working memory storage and manipulation. Increased complexity in cytoarchitecture, connectivity, dendritic spine density, and receptor expression additionally reveal a sharper hierarchical organization in primate cortex. Together, we argue that these primate specializations permit separable deconstruction and selective reconstruction of representations, which is essential to higher cognition.

Design of sports achievement prediction system based on U-net convolutional neural network in the context of machine learning

Sports plays a pivotal role in national development. To accurately predict college students' sports performance and motivate them to improve their physical fitness, this study constructs a sports achievement prediction system by using a U-Net Convolutional Neural Network (CNN) in machine learning. Firstly, the current state of physical education teachers’ instructional proficiency is investigated and analyzed to identify existing problems. Secondly, an improved U-Net-based sports achievement prediction system is proposed. This method enhances the utilization and propagation of network features by incorporating dense connections, thus addressing gradient disappearance issues. Simultaneously, an improved mixed loss function is introduced to alleviate class imbalance. Finally, the effectiveness of the proposed system is validated through testing, demonstrating that the improved U-Net CNN algorithm yields superior results. Specifically, the prediction accuracy of the improved network for sports performance surpasses that of the original U-Net by 4.22 % and exceeds that of DUNet by 5.22 %. Compared with other existing prediction networks, the improved U-Net CNN model exhibits a superior achievement prediction ability. Consequently, the proposed system enhances teaching and learning efficiency and offers insights into applying artificial intelligence technology to smart classroom development.

Succession of tissue microbial community during oat developmental

Investigating oat tissue microflora during its different developmental stages is necessary for understanding its growth and anti-disease mechanism. In this study, 16S rDNA and ITS (Internally Transcribed Spacer) high-throughput sequencing technology were used to explore the microflora diversity of oat tissue. Twenty-seven samples of leaves, stems, and roots from three developmental stages, namely the seedling stage (SS), jointing stage (JS), and maturity stage (MS), underwent sequencing analysis. The analysis showed that 6480 operational taxonomic units (OTUs) were identified in the examined samples, of which 1698 were fungal and 4782 were bacterial. Furthermore, 126 OTUs were shared by fungi, mainly Ascomycota, Basidiomycota, and Mucoromycota at the phylum level, and 39 OTUs were shared by bacteria, mainly Actinobacteriota and Proteobacteria at the phylum level. The microbial diversity of oat tissue in the three developmental stages showed differences, and the α-diversity of the bacteria and β-diversity of the bacteria and fungi in the roots were higher than those of the stems and leaves. Among the bacteria species, Thiiopseudomonas, Rikenellaceae RC9 gut group, and Brevibacterium were predominant in the leaves, MND1 was predominant in the roots, and Lactobacillus was predominant in the stems. Moreover, Brevibacterium maintained a stable state at all growth stages. In the fungal species, Phomatospora was dominant in the leaves, Kondoa was dominant in the roots, and Pyrenophora was dominant in the stems. All species with a high abundance were related to the growth process of oats and antagonistic bacteria. Furthermore, connection modules were denser in bacterial than in fungal populations. The samples were treated with superoxide dismutase and peroxidase. There were 42 strains associated with SOD (Superoxide dismutase), 60 strains associated with POD (Peroxidase), and 38 strains in total, which much higher than fungi. The network analysis showed that bacteria might have more dense connection modules than fungi, The number of bacterial connections to enzymes were much higher than that of fungi. Furthermore, these results provide a basis for further mechanistic research.

Transfer Learning-Based Hyperspectral Image Classification Using Residual Dense Connection Networks.

The extraction of effective classification features from high-dimensional hyperspectral images, impeded by the scarcity of labeled samples and uneven sample distribution, represents a formidable challenge within hyperspectral image classification. Traditional few-shot learning methods confront the dual dilemma of limited annotated samples and the necessity for deeper, more effective features from complex hyperspectral data, often resulting in suboptimal outcomes. The prohibitive cost of sample annotation further exacerbates the challenge, making it difficult to rely on a scant number of annotated samples for effective feature extraction. Prevailing high-accuracy algorithms require abundant annotated samples and falter in deriving deep, discriminative features from limited data, compromising classification performance for complex substances. This paper advocates for an integration of advanced spectral-spatial feature extraction with meta-transfer learning to address the classification of hyperspectral signals amidst insufficient labeled samples. Initially trained on a source domain dataset with ample labels, the model undergoes transference to a target domain with minimal samples, utilizing dense connection blocks and tree-dimensional convolutional residual connections to enhance feature extraction and maximize spatial and spectral information retrieval. This approach, validated on three diverse hyperspectral datasets-IP, UP, and Salinas-significantly surpasses existing classification algorithms and small-sample techniques in accuracy, demonstrating its applicability to high-dimensional signal classification under label constraints.

LFMNet: a lightweight model for identifying leaf diseases of maize with high similarity.

Maize leaf diseases significantly impact yield and quality. However, recognizing these diseases from images taken in natural environments is challenging due to complex backgrounds and high similarity of disease spots between classes.This study proposes a lightweight multi-level attention fusion network (LFMNet) which can identify maize leaf diseases with high similarity in natural environment. The main components of LFMNet are PMFFM and MAttion blocks, with three key improvements relative to existing essential blocks. First, it improves the adaptability to the change of maize leaf disease scale through the dense connection of partial convolution with different expansion rates and reduces the parameters at the same time. The second improvement is that it replaces a adaptable pooling kernel according to the size of the input feature map on the original PPA, and the convolution layer to reshape to enhance the feature extraction of maize leaves under complex background. The third improvement is that it replaces different pooling kernels to obtain features of different scales based on GMDC and generate feature weighting matrix to enhance important regional features. Experimental results show that the accuracy of the LFMNet model on the test dataset reaches 94.12%, which is better than the existing heavyweight networks, such as ResNet50 and Inception v3, and lightweight networks such as DenseNet 121,MobileNet(V3-large) and ShuffleNet V2. The number of parameters is only 0.88m, which is better than the current mainstream lightweight network. It is also effective to identify the disease types with similar disease spots in leaves.

A Super-Resolution Reconstruction Model for Remote Sensing Image Based on Generative Adversarial Networks

In current times, reconstruction of remote sensing images using super-resolution is a prominent topic of study. Remote sensing data have a complex spatial distribution. Compared with natural pictures, remote sensing pictures often contain subtler and more complicated information. Most super-resolution reconstruction algorithms cannot restore all the information contained in remote sensing images when reconstructing them. The content of some areas in the reconstructed images may be too smooth, and some areas may even have color changes, resulting in lower quality reconstructed images. In response to the problems presenting in current reconstruction algorithms about super-resolution, this article proposes the SRGAN-MSAM-DRC model (SRGAN model with multi-scale attention mechanism and dense residual connection). This model roots in generative adversarial networks and incorporates multi-scale attention mechanisms and dense residual connections into the generator. Furthermore, residual blocks are incorporated into the discriminator. We use some remote sensing image datasets of real-world data to evaluate this model, and the results indicate the SRGAN-MSAM-DRC model has shown enhancements in three evaluation metrics for reconstructed images about super-resolution. Compared to the basic SRGAN model, the SSIM (structural similarity), PSNR (peak signal-to-noise ratio), and IE (image entropy) increase by 5.0%, 4.0%, and 4.1%, respectively. From the results, we know the quality of the reconstructed images of remote sensing using the SRGAN-MSAM-DRC model is better than basic SRGAN model, and verifies that the model has good applicability and performance in reconstruction of remote sensing images using super-resolution.

DIRECTION: Deep cascaded reconstruction residual-based feature modulation network for fast MRI reconstruction

Deep cascaded networks have been extensively studied and applied to accelerate Magnetic Resonance Imaging (MRI) and have shown promising results. Most existing works employ a large cascading number for the sake of superior performances. However, due to the lack of proper guidance, the reconstruction performance can easily reach a plateau and even face degradation if simply increasing the cascading number. In this paper, we aim to boost the reconstruction performance from a novel perspective by proposing a parallel architecture called DIRECTION that fully exploits the guiding value of the reconstruction residual of each subnetwork. Specifically, we introduce a novel Reconstruction Residual-Based Feature Modulation Mechanism (RRFMM) which utilizes the reconstruction residual of the previous subnetwork to guide the next subnetwork at the feature level. To achieve this, a Residual Attention Modulation Block (RAMB) is proposed to generate attention maps using multi-scale residual features to modulate the image features of the corresponding scales. Equipped with this strategy, each subnetwork within the cascaded network possesses its unique optimization objective and emphasis rather than blindly updating its parameters. To further boost the performance, we introduce the Cross-Stage Feature Reuse Connection (CSFRC) and the Reconstruction Dense Connection (RDC), which can reduce information loss and enhance representative ability. We conduct sufficient experiments and evaluate our method on the fastMRI knee dataset using multiple subsampling masks. Comprehensive experimental results show that our method can markedly boost the performance of cascaded networks and significantly outperforms other compared state-of-the-art methods quantitatively and qualitatively.