Existing Network Architecture Research Articles

Lightweight camouflaged object detection model based on multilevel feature fusion

The intrinsic similarity between camouflaged objects and background environment impedes the automatic detection/segmentation of camouflaged objects, and novel network architectures for deep learning are promising to overcome this challenge and improve detection accuracy. However, these existing network architectures for distinguishing between camouflaged objects and their backgrounds do not account for the constraint of detection speed, which results in high computational complexity and the inability to meet the requirements of rapid detection. Therefore, based on the human visual system, this study proposes a single-stage lightweight camouflage object detection network using multilevel feature fusion, integrating features of various feature layers and receptive field sizes. Using three benchmark datasets for normal camouflaged objects, the lightweight network (LINet) model demonstrated an accuracy superior to those of six existing mainstream camouflaged object detection methods. Its detection speed, 126.3 frames per second, is significantly higher than those of the existing mainstream methods, enabling rapid detection with a maximum increase of 187.62%. The accuracy of LINet is the minimum and maximum for Resnet101 and Resnet152, respectively. These findings pave the way for diverse applications of camouflaged target detection algorithms.

Deep reinforcement learning for base station switching scheme with federated LSTM‐based traffic predictions

AbstractTo meet increasing traffic requirements in mobile networks, small base stations (SBSs) are densely deployed, overlapping existing network architecture and increasing system capacity. However, densely deployed SBSs increase energy consumption and interference. Although these problems already exist because of densely deployed SBSs, even more SBSs are needed to meet increasing traffic demands. Hence, base station (BS) switching operations have been used to minimize energy consumption while guaranteeing quality‐of‐service (QoS) for users. In this study, to optimize energy efficiency, we propose the use of deep reinforcement learning (DRL) to create a BS switching operation strategy with a traffic prediction model. First, a federated long short‐term memory (LSTM) model is introduced to predict user traffic demands from user trajectory information. Next, the DRL‐based BS switching operation scheme determines the switching operations for the SBSs using the predicted traffic demand. Experimental results confirm that the proposed scheme outperforms existing approaches in terms of energy efficiency, signal‐to‐interference noise ratio, handover metrics, and prediction performance.

Understanding Human-Computer Interactions in Restricted Clinical Environments.

Conducting research on human-computer interaction and information retrieval requires unobtrusive observations within existing network architectures. Most of the available tools are not suitable to be applied within restricted clinical systems. The specific requirements hinder analysis of the human factors in health sciences. We identified extensions for popular web browsers as a suitable way to conduct studies in highly regulated environments. Considering the specialized requirements and an adequate level of transparency for the recorded clinician, we developed an open-source Web Extension compatible with major web browsers. We identified the challenges associated with the specific tool and are preparing its use to understand clinical reasoning in personalized oncology.

CODE: Computation Offloading in D2D-Edge System for Video Streaming

Video traffic over the Internet is increasing rapidly, reaching up to 82% of the total Internet traffic by 2022. This enormous growth of video traffic which also include immersive video content (augmented reality and virtual reality streaming) is challenging for the existing network architectures. Mobile edge computing (MEC) is arising as a promising solution to enhance the user experience. However, edge servers have usually limited computational capabilities and may not cope with unplanned sudden spikes in traffic from live streaming of large events or when a video goes viral, known as the <italic xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">thundering herd problem</i> . To improve the edge resource utilization, cooperation among edge servers has been proposed earlier. However, such cooperation is not useful in single-server scenario or when the neighboring servers are also fully utilized. Thus, in this article, we propose a novel D2D-MEC collaborative framework called as CODE (Computation Offloading in D2D-Edge) system in which an edge server offloads its computations to the distributed user devices (downlink). Two distinct schemes, i.e., maximal offloading with delay constraint (MOD) and minimum delay offloading (MDO) and a lightweight heuristic, are proposed to solve the computation offloading problem. Our simulation results report that the MOD scheme reduces the network delay up to 65%, whereas the MDO scheme reduces the edge computation load up to 83%.

A Novel Approach to Optimize Memory Reconstruction Using Joint Multimodal Networks

Memories play a crucial role in the human life experience. Whether they are used to make decisions or understand current situations, memories influence the past, present, and future. However, some memories are unreliable; in the court of law, the false memories of victims can cause inaccurate verdicts. To reconstruct human memories directly from the brain, recent studies have used electroencephalographic (EEG) brain signals in machine learning frameworks like generative adversarial networks (GANs), which generates new data from old ones. However, traditional GANs tend to produce the same images, an issue called mode collapse. Therefore, a conditional variational autoencoder-generative adversarial network (CVAE-GAN), which had not been used before in memory reconstruction, was developed to address GAN failures by jointly-training (1) an encoder and (2) GAN. CVAE-GAN correlated the MindBigData dataset’s EEG brain signals with the corresponding ImageNet dataset images and produced new memories of participants, such as pandas, humans, and fish. CVAE-GAN’s Inception Score, which marked how diverse and distinct its generated images were, was 1.00 on average. The number of floating operations, or FLOPS, was 102 Gigaflops, which was less than the traditional GAN’s 198 Gigaflops. While limitations with time, computational memory, and mode collapse prevented the CVAE-GAN from recreating accurate memories, it still generated distinct image colors and general features. Future studies can build on existing network architectures and include more homogeneity in datasets. Ultimately, CVAE-GAN has the potential to advance new understandings of the brain and could elevate the memory reconstruction field.

LRPRNet: Lightweight Deep Network by Low-Rank Pointwise Residual Convolution.

Deep learning has become popular in recent years primarily due to powerful computing devices such as graphics processing units (GPUs). However, it is challenging to deploy these deep models to multimedia devices, smartphones, or embedded systems with limited resources. To reduce the computation and memory costs, we propose a novel lightweight deep learning module by low-rank pointwise residual (LRPR) convolution, called LRPRNet. Essentially, LRPR aims at using a low-rank approximation in pointwise convolution to further reduce the module size while keeping depthwise convolutions as the residual module to rectify the LRPR module. This is critical when the low-rankness undermines the convolution process. Moreover, our LRPR is quite general and can be directly applied to many existing network architectures such as MobileNetv1, ShuffleNetv2, MixNet, and so on. Experiments on visual recognition tasks, including image classification and face alignment on popular benchmarks, show that our LRPRNet achieves competitive performance but with a significant reduction of Flops and memory cost compared to the state-of-the-art deep lightweight models.

Two Revised Deep Neural Networks and Their Applications in Quantitative Analysis Based on Near-Infrared Spectroscopy

Small data sets make developing calibration models using deep neural networks difficult because it is easy to overfit the system. We developed two deep neural network architectures by revising two existing network architectures: the U-Net and the attention mechanism. The major changes were to use 1D convolutional layers to replace the fully connected layers. We also designed and combined average pooling and maximum pooling in our revised networks, respectively. We applied these revised network architectures to three publicly available data sets and the resulting calibration models can generate acceptable results for general quantitative analysis. It also generated rather good results for data sets that concern calibration transfer. It demonstrates that constructing network architectures by properly revising existing successful network architectures may provide additional choices in the exploration of the application of deep neural network in analytical chemistry.

Deep Learning for Ultrasound Speed-of-Sound Reconstruction: Impacts of Training Data Diversity on Stability and Robustness

Ultrasound b-mode imaging is a qualitative approach and diagnostic quality strongly depends on operators’ training and experience. Quantitative approaches can provide information about tissue properties; therefore, can be used for identifying various tissue types, e.g., speed-of-sound in the tissue can be used as a biomarker for tissue malignancy, especially in breast imaging. Recent studies showed the possibility of speed-of-sound reconstruction using deep neural networks that are fully trained on simulated data. However, because of the ever-present domain shift between simulated and measured data, the stability and performance of these models in real setups are still under debate. In prior works, for training data generation, tissue structures were modeled as simplified geometrical structures which does not reflect the complexity of the real tissues. In this study, we proposed a new simulation setup for training data generation based on Tomosynthesis images. We combined our approach with the simplified geometrical model and investigated the impacts of training data diversity on the stability and robustness of an existing network architecture. We studied the sensitivity of the trained network to different simulation parameters, e.g., echogenicity, number of scatterers, noise, and geometry. We showed that the network trained with the joint set of data is more stable on out-of-domain simulated data as well as measured phantom data.

SGAT: Snapshot-guided adversarial training of neural networks

It is well known that ensembles of different models can achieve better performance. Unfortunately, training multiple deep neural networks is time-consuming and expensive. Also, they require more space for storage and more computation for inference, making them unsuitable for applications with limited resources, such as mobile and embedded devices. To address these two fundamental problems in ensemble learning, we propose a novel method called Snapshot-Guided Adversarial Training (SGAT) to automatically accumulate and distill knowledge from the training, where knowledge from the earlier training iterations of the network is distilled, and transferred into its late training iterations via an adversarial learning strategy. To accumulate knowledge from the training process, we employ a cyclic annealing schedule and take a model snapshot at the end of each training interval. Furthermore, we use a shared discriminator to encourage the distillation process. The main advantages of our method are: (1) We can directly train a network with its free snapshots for knowledge distillation, instead of heavily depending on pre-trained models; (2) the inference cost remains the same as using a single model after finishing knowledge transferring; (3) SGAT is a general method and can be applied to the existing network architectures. Our extensive experiments show that SGAT consistently outperforms the standing training method with a clear margin. For example, with the same training budget, it achieves 2.86% more accuracy on average than the baseline when training MobileNet-V2 from scratch on CIFAR-100. Meanwhile, SGAT also achieves better performance in most cases than the existing ensemble method and knowledge distillation. For example, to train MobileNet-V2 from scratch on ImageNet, it gets 0.7% more accuracy than snapshot ensembles, and 0.93% more accuracy than snapshot distillation. More importantly, our accumulated learning strategy makes SGAT achieve much better performance when we increase training time. For example, compared with the standard training method for MobileNet-V2, it gets 3.13% more accuracy on ImageNet.

An industry perspective for algorithm performance, testing, and validation

It is well known that ensembles of different models can achieve better performance. Unfortunately, training multiple deep neural networks is time-consuming and expensive. Also, they require more space for storage and more computation for inference, making them unsuitable for applications with limited resources, such as mobile and embedded devices. To address these two fundamental problems in ensemble learning, we propose a novel method called Snapshot-Guided Adversarial Training (SGAT) to automatically accumulate and distill knowledge from the training, where knowledge from the earlier training iterations of the network is distilled, and transferred into its late training iterations via an adversarial learning strategy. To accumulate knowledge from the training process, we employ a cyclic annealing schedule and take a model snapshot at the end of each training interval. Furthermore, we use a shared discriminator to encourage the distillation process. The main advantages of our method are: (1) We can directly train a network with its free snapshots for knowledge distillation, instead of heavily depending on pre-trained models; (2) the inference cost remains the same as using a single model after finishing knowledge transferring; (3) SGAT is a general method and can be applied to the existing network architectures. Our extensive experiments show that SGAT consistently outperforms the standing training method with a clear margin. For example, with the same training budget, it achieves 2.86% more accuracy on average than the baseline when training MobileNet-V2 from scratch on CIFAR-100. Meanwhile, SGAT also achieves better performance in most cases than the existing ensemble method and knowledge distillation. For example, to train MobileNet-V2 from scratch on ImageNet, it gets 0.7% more accuracy than snapshot ensembles, and 0.93% more accuracy than snapshot distillation. More importantly, our accumulated learning strategy makes SGAT achieve much better performance when we increase training time. For example, compared with the standard training method for MobileNet-V2, it gets 3.13% more accuracy on ImageNet.

SiamLight: lightweight networks for object tracking via attention mechanisms and pixel-level cross-correlation

Despite Siamese-based trackers have achieved great success in recent years, researchers have focused more on the accuracy of trackers than their complexity, which leads to their inapplicability in some scenarios, and the real-time speed can be greatly limited. In this work, we propose a lightweight network method called SiamLight for object tracking. MobileNet-V3 is selected as the backbone network. The PG-corr module is added as the feature fusion module, a strategy that decomposes the template feature into spatial and channel kernels, reducing the matching regions and suppressing the effect of similar interference. In addition, we also add the CSM module, which carries out attention to the channel and spatial simultaneously. CSM module not only reduces the number of parameters but also ensures that it can be integrated into existing network architectures as a plug-and-play module. Finally, multiple separable convolution blocks are added to the classification and regression branches to meet our lightweight parameters and Flops requirements. The experiments on LaSOT, VOT2018, VOT2019, OTB100, and UAV123 benchmarks show that the method has fewer Flops and parameters than state-of-the-art trackers.

Kernel-based feature aggregation framework in point cloud networks

Various effective deep networks have been developed for analysis of 3D point clouds. One key step in these networks is to aggregate the features of orderless points into a compact representation for the cloud. As a typical order-invariant aggregation method, max-pooling has been widely applied. However, while enjoying simplicity and high efficiency, max-pooling does not fully exploit the feature information since it not only ignores the non-maximum elements in each feature dimension but also neglects the interactions between different dimensions. These drawbacks of max-pooling motivate us to explore advanced feature aggregation methods for 3D point cloud analysis. The desired advanced method should be capable of modeling richer information between the point features than max-pooling, and, at the same time, it can readily replace max-pooling without the need to modify other parts of the existing network architecture. To this end, this paper proposes a novel kernel-based feature aggregation framework for 3D point cloud analysis for the first time. The proposed method effectively considers all the elements in each dimension and models the nonlinear interactions between feature dimensions as complementary information to max-pooling. In addition, it is a plug-in module that can be integrated to many common networks as a replacement of max-pooling. Comprehensive experiments are conducted to demonstrate the consistently superior performance and high generality of the proposed method over max-pooling. Specifically, the proposed kernel-based feature aggregation framework consistently improves the max-pooling with three representative backbones of PointNet, DGCNN and PCT across four 3D point cloud based analysis tasks, including supervised 3D object classification, 3D part segmentation, indoor semantic segmentation and one additional unsupervised place retrieval task. Especially, it shows remarkable performance improvement over max-pooling in the unsupervised retrieval task, demonstrating its advantage in forming 3D point cloud representation.

PerimetryNet: A multiscale fine grained deep network for three‐dimensional eye gaze estimation using visual field analysis

AbstractThree‐dimensional gaze estimation aims to reveal where a person is looking, which plays an important role in identifying users' point‐of‐interest in terms of the direction, attention and interactions. Appearance‐based gaze estimation methods could provide relatively unconstrained gaze tracking from commodity hardware. Inspired by medical perimetry test, we have proposed a multiscale framework with visual field analysis branch to improve estimation accuracy. The model is based on the feature pyramids and predicts vision field to help gaze estimation. In particular, we analysis the effect of the multiscale component and the visual field branch on challenging benchmark datasets: MPIIGaze and EYEDIAP. Based on these studies, our proposed PerimetryNet significantly outperforms state‐of‐the‐art methods. In addition, the multiscale mechanism and visual field branch can be easily applied to existing network architecture for gaze estimation. Related code would be available at public repository https://github.com/gazeEs/PerimetryNet.

Role of Cyber Twin Mobile Networks in Digital Communication

Due to the unprecedented increase in the utilization of Internet of Things (IoT) technology, there is a huge increase in network/internet traffic due to which the existing network architectures are facing number of challenges in terms of reliability, availability, scalability, portability and expandability. In order to deal with the challenges, the cyber-twin technology has been recently proposed to assist in establishing a strong communication network in terms of digital asset management, network data maintenance etc. Recently, cyber-twin has become a well-established environment for real-world applications ranging from manufacturing, healthcare to smart city applications. The proposed study reviews the concept of cyber-twin mobile networks along with a representation and efficient utilization of cyber-twin networks and its impact in the digital world. Further, different state-of-the-art cyber-twin architectures are compared to show how these architectures outperform the traditional peer-to-peer networks in order to make the futuristic networks safer, secured, adaptable, portable and reliable.

Enriching Point Clouds with Implicit Representations for 3D Classification and Segmentation

Continuous implicit representations can flexibly describe complex 3D geometry and offer excellent potential for 3D point cloud analysis. However, it remains challenging for existing point-based deep learning architectures to leverage the implicit representations due to the discrepancy in data structures between implicit fields and point clouds. In this work, we propose a new point cloud representation by integrating the 3D Cartesian coordinates with the intrinsic geometric information encapsulated in its implicit field. Specifically, we parameterize the continuous unsigned distance field around each point into a low-dimensional feature vector that captures the local geometry. Then we concatenate the 3D Cartesian coordinates of each point with its encoded implicit feature vector as the network input. The proposed method can be plugged into an existing network architecture as a module without trainable weights. We also introduce a novel local canonicalization approach to ensure the transformation-invariance of encoded implicit features. With its local mechanism, our implicit feature encoding module can be applied to not only point clouds of single objects but also those of complex real-world scenes. We have validated the effectiveness of our approach using five well-known point-based deep networks (i.e., PointNet, SuperPoint Graph, RandLA-Net, CurveNet, and Point Structuring Net) on object-level classification and scene-level semantic segmentation tasks. Extensive experiments on both synthetic and real-world datasets have demonstrated the effectiveness of the proposed point representation.

Optimal assessment of smart grid based photovoltaic cell operational parameters using simulated annealing

With the ever-increasing load demand throughout the globe, natural renewable resources integrated into the existing network architecture for sustainable energy production are gaining considerable significance. Photovoltaic (PV) generation systems is one such technique to deal with the worldwide challenge for achieving green energy and low carbon footprint while simultaneously providing emission free electrical power from solar radiations. In this paper, we consider smart grid architecture connecting the end-users and the utility power plant with solar energy sources through an effective power optimization system. Multiple performance criteria associated with solar cell operation are evaluated and analyzed using the simulated annealing algorithm. These objectives considered for optimization include the cell saturation current, photo-generated current, material band gap, cell temperature, annualized life cycle cost, fill factor and cell efficiency. The formulated optimization conditions are specified in terms of two independent variables of cell ambient temperature and cell illumination. Moreover, the adaption of distinct values of short circuit current coefficients on the light originated current is measured. Through extensive simulation experiments, two disparate annealing procedures of fast annealing and Boltzmann annealing are applied coupled with three categories of temperature update schemes, viz. exponential, logarithmic and linear.

Interpretability-Guided Inductive Bias For Deep Learning Based Medical Image.

Deep learning methods provide state of the art performance for supervised learning based medical image analysis. However it is essential that trained models extract clinically relevant features for downstream tasks as, otherwise, shortcut learning and generalization issues can occur. Furthermore in the medical field, trustability and transparency of current deep learning systems is a much desired property. In this paper we propose an interpretability-guided inductive bias approach enforcing that learned features yield more distinctive and spatially consistent saliency maps for different class labels of trained models, leading to improved model performance. We achieve our objectives by incorporating a class-distinctiveness loss and a spatial-consistency regularization loss term. Experimental results for medical image classification and segmentation tasks show our proposed approach outperforms conventional methods, while yielding saliency maps in higher agreement with clinical experts. Additionally, we show how information from unlabeled images can be used to further boost performance. In summary, the proposed approach is modular, applicable to existing network architectures used for medical imaging applications, and yields improved learning rates, model robustness, and model interpretability.

Architecture, Ability, and Adaptability of Recursive Internetworking Architecture - A Review

: The ever-increasing dependency of the utilities on networking brought several cyber vulnerabilities and burdened them with dynamic networking demands like QoS, multihoming, and mobility. As the existing network was designed without security in context, it poses several limitations in mitigating the unwanted cyber threats and struggling to provide an integrated solution for the novel networking demands. These limitations resulted in the design and deployment of various add-on protocols that made the existing network architecture a patchy and complex network. The proposed work introduces one of the future internet architectures, which seem to provide abilities to mitigate the above limitations. Recursive internetworking architecture (RINA) is one of the future internets and appears to be a reliable solution with its promising design features. RINA extended inter-process communication to distributed inter-process communication and combined it with recursion. RINA offered unique inbuilt security and the ability to meet novel networking demands with its design. It has also provided integration methods to make use of the existing network infrastructure. The present work reviews the unique architecture, abilities, and adaptability of RINA based on various research works of RINA. The contribution of this article is to expose the potential of RINA in achieving efficient networking solutions among academia and industry.

An Improved YOLOX Algorithm for Forest Insect Pest Detection

A large number of insect pests in the forest will seriously affect the construction of forest resources and agriculture in China. In this regard, in order to deeply understand and analyze the existing forest pest detection technology, it is found that it cannot meet practical needs. In order to prevent the harm caused by forest pests, it is necessary to correctly identify the types of pests and take targeted control measures. Therefore, this paper proposes a forest pest detection algorithm based on improved YOLOX. Firstly, aiming at the problem that there are few image data of real deep forest pests in the wild, we use Mosaic, Mixup, and random erasure data enhancement to preprocess the images. Secondly, in order to extract fine-grained features, shallow information is introduced into the existing network architecture, and a two-way cross-scale feature fusion mechanism is adopted. Finally, the improved YOLOX algorithm proposed in this paper has achieved the best results on the public forest pest dataset IP102.

Improving Deep Mutual Learning via Knowledge Distillation

Knowledge transfer has become very popular in recent years, and it is either based on a one-way transfer method used with knowledge distillation or based on a two-way knowledge transfer implemented by deep mutual learning, while both of them adopt a teacher–student paradigm. A one-way based method is more simple and compact because it only involves an untrained low-capacity student and a high-capacity teacher network in the knowledge transfer process. In contrast, a two-way based method requires more training costs because it involves two or more low-cost network capacities from scratch simultaneously to obtain better accuracy results for each network. In this paper, we propose two new approaches, namely full deep distillation mutual learning (FDDML) and half deep distillation mutual learning (HDDML), and improve convolutional neural network performance. These approaches work with three losses by using variations of existing network architectures, and the experiments have been conducted on three public benchmark datasets. We test our method on some existing KT task methods, showing its performance over related methods.