Network In Network Research Articles

Growth-adaptive distillation compressed fusion model for network traffic identification based on IoT cloud-edge collaboration

The development of the Internet of Things (IoT) has led to the rapid growth of the types and number of connected devices and has generated large amounts of complex and diverse traffic data. Traffic identification on edge servers solves the real-time and privacy requirements of IoT management and has attracted much attention, but still faces several problems: (1) traditional machine learning (ML) models rely on artificially constructed features, and the existing deep learning (DL) traffic identification models have reached their performance limit; and (2) insufficient computing resources of edge servers limit the possible improvement in the performance of deep learning models by increasing the number of parameters and structural complexity. To address these issues, we propose a lightweight fusion model. First, the Network-in-Network (NiN) model and Random Forest (RF) model are used on the cloud server to construct a traffic identification fusion model. The excellent representation extraction capability of the NiN compensates for the RF’s dependence on manual feature extraction, and its modular structure is suitable for the subsequent model compression operations. Then, the NiN was distilled. We propose Growth-Adaptive Distillation to lightweight the NiN model, which can reduce the operation of manually adjusting the structure of the student model and ensure the efficiency and low power consumption of the fusion model deployment. In addition, both the RF in the cloud and the distilled NiN are deployed on the edge server. Comparisons with multiple algorithms on two network traffic datasets show that the proposed model achieves state-of-the-art performance while ensuring the use of minimal computational resources.

Identification of Passion Fruit Nutrients for Elderly People Using Network in Network Architecture: An Empirical Study in Thailand

The growing elderly population has led to a rise in health issues, particularly chronic diseases. Passion fruits contain numerous nutrients that may help in the treatment of chronic diseases. However, specific recommendations for daily passion fruit nutrient intake for the elderly are currently lacking in the literature. This research aimed to identify passion fruit groups and to suggest the appropriate daily passion fruit nutrient intake for elderly people using network in network (NiN) architecture. This research demonstrates that the NiN model can be effectively applied to identify passion fruit groups for the elderly. It is more efficient than other convolutional neural network (CNN) architectures. The results show that NiN can correctly identify passion fruit groups and suggest the appropriate amount of nutrient intake for the elderly, achieving + 96.76% accuracy in the training dataset and 95.89% accuracy in the validation dataset, surpassing 84.6% accuracy achieved by EaglAI. Sensitivity analysis of the NiN model using mean absolute error (MAE) for geometric transformations revealed consistent training image results and model robustness. This research benefits elderly people with chronic diseases by providing tailored recommendations for daily passion fruit intake, based on the analysis of sugar nutrients using the NiN model.

Encrypted Network Traffic Classification Using Deep and Parallel Network-In-Network Models

Network traffic classification analyzes received data packets to identify distinct application or traffic kinds. This research describes a neural network model that uses deep and parallel network-in-network (NIN) architectures to classify encrypted network data. In comparison to typical convolutional neural networks (CNN), NIN uses a micro network after each convolution layer to improve local modeling. Furthermore, NIN uses global average pooling instead of traditional fully connected layers before final classification, resulting in a considerable reduction in the number of model parameters. Our suggested solution uses deep NIN models with several MLP convolutional layers to map fixed-length packet vectors to application or traffic labels. Furthermore, a parallel decision method is created to build two sub-networks to process packet headers and packet bodies separately, taking into account that they may include different types of evidence for classification. Our investigations on the ''ISCX VPN-nonVPN'' encrypted traffic dataset demonstrate that NIN models can achieve a better balance between classification accuracy and model complexity than standard CNNs. The parallel decision technique can increase the accuracy of a single NIN model for classifying encrypted network data. Finally, the test set F1 scores of 0.983 and 0.985 are obtained for traffic characterisation and application identification, respectively

Image-based 3D reconstruction and permeability modelling of rock using enhanced interpretable deep residual learning

The study presents a novel deep residual learning framework with interpretability for the prediction of rock permeability. The proposed framework, termed ResNet-NiN integrates a 3D deep residual network (ResNet) with network in network (NiN) connections. A total of 1331 datasets were generated using voxel segmentation of DRP samples obtained from actual Berea Sandstone and subsequently, a percolation simulation was performed. The learning rate decay strategy of Cosine Annealing with warmup restarts is used to train the predictive model and prevent the predicted permeability from being trapped in the local optima. To improve the trust and comprehensibility of the model and visualize the ResNet-NiN architecture, we established a compound approach that leverages the internal gradient change of the neural network into the deep learning framework. This integrated approach encompasses the analysis of Integrated Gradients (IG) from the input perspective and Gradient-weighted Class Activation Mapping (Grad-CAM) from the output perspective. The experimental findings demonstrate that the developed model exhibits superiority over the typically employed 3D convolutional neural network architecture in terms of both accuracy and stability. The unified interpretability technique offers a coherent explanation for the provided sample, validating the model's capability and elucidating the factors and relationships underlying the model's prediction.

Steering a Robotic Wheelchair Based on Voice Recognition System Using Convolutional Neural Networks

Many wheelchair people depend on others to control the movement of their wheelchairs, which significantly influences their independence and quality of life. Smart wheelchairs offer a degree of self-dependence and freedom to drive their own vehicles. In this work, we designed and implemented a low-cost software and hardware method to steer a robotic wheelchair. Moreover, from our method, we developed our own Android mobile app based on Flutter software. A convolutional neural network (CNN)-based network-in-network (NIN) structure approach integrated with a voice recognition model was also developed and configured to build the mobile app. The technique was also implemented and configured using an offline Wi-Fi network hotspot between software and hardware components. Five voice commands (yes, no, left, right, and stop) guided and controlled the wheelchair through the Raspberry Pi and DC motor drives. The overall system was evaluated based on a trained and validated English speech corpus by Arabic native speakers for isolated words to assess the performance of the Android OS application. The maneuverability performance of indoor and outdoor navigation was also evaluated in terms of accuracy. The results indicated a degree of accuracy of approximately 87.2% of the accurate prediction of some of the five voice commands. Additionally, in the real-time performance test, the root-mean-square deviation (RMSD) values between the planned and actual nodes for indoor/outdoor maneuvering were 1.721 × 10−5 and 1.743 × 10−5, respectively.

Development of Voice Control Algorithm for Robotic Wheelchair Using MIN and LSTM Models

In this work, we developed and implemented a voice control algorithm to steer smart robotic wheelchairs (SRW) using the neural network technique. This technique used a network in network (NIN) and long short-term memory (LSTM) structure integrated with a built-in voice recognition algorithm. An Android Smartphone application was designed and configured with the proposed method. A Wi-Fi hotspot was used to connect the software and hardware components of the system in an offline mode. To operate and guide SRW, the design technique proposed employing five voice commands (yes, no, left, right, no, and stop) via the Raspberry Pi and DC motors. Ten native Arabic speakers trained and validated an English speech corpus to determine the method’s overall effectiveness. The design method of SRW was evaluated in both indoor and outdoor environments in order to determine its time response and performance. The results showed that the accuracy rate for the system reached 98.2% for the five-voice commends in classifying voices accurately. Another interesting finding from the real-time test was that the root-mean-square deviation (RMSD) for indoor/outdoor maneuvering nodes was 2.2*10–5 (for latitude), while that for longitude coordinates was a whopping 2.4*10–5 (for latitude).

Combination of gated recurrent unit and Network in Network for underwater acoustic target recognition

Underwater acoustic target recognition is an important part of underwater acoustic signal processing and an important technical support for underwater acoustic information acquisition and underwater acoustic information confrontation. Taking into account that the gated recurrent unit (GRU) has an internal feedback mechanism that can reflect the temporal correlation of underwater acoustic target features, a model with gated recurrent unit and Network in Network (NIN) is proposed to recognize underwater acoustic targets in this paper. The proposed model introduces NIN to compress the hidden states of GRU while retaining the original timing characteristics of underwater acoustic target features. The higher recognition rate and faster calculation speed of the proposed model are demonstrated with experiments for raw underwater acoustic signals comparing with the multi-layer stacked GRU model.

Enhanced Diagnosis of Pneumothorax with an Improved Real-Time Augmentation for Imbalanced Chest X-rays Data Based on DCNN.

Pneumothorax is a common pulmonary disease that can lead to dyspnea and can be life-threatening. X-ray examination is the main means to diagnose this disease. Computer-aided diagnosis of pneumothorax on chest X-ray, as a prerequisite for a timely cure, has been widely studied, but it is still not satisfactory to achieve highly accurate results. In this paper, an image classification algorithm based on the deep convolutional neural network (DCNN) is proposed for high-resolution medical image analysis of pneumothorax X-rays, which features a Network In Network (NIN) for cleaning the data, random histogram equalization data augmentation processing, and a DCNN. The experimental results indicate that the proposed method can effectively increase the correct diagnosis rate of pneumothorax, and the Area under Curve (AUC) of the test verified in the experiment is 0.9844 on ZJU-2 test data and 0.9906 on the ChestX-ray14, respectively. In addition, a large number of atmospheric pleura samples are visualized and analyzed based on the experimental results and in-depth learning characteristics of the algorithm. The analysis results verify the validity of feature extraction for the network. Combined with the results of these two aspects, the proposed X-ray image processing algorithm can effectively improve the classification accuracy of pneumothorax photographs.

A robust and fast multispectral pedestrian detection deep network

Multispectral pedestrian detection is a difficult task, especially with pedestrian images of different sizes. In most convolutional neural network (CNN) models, the shared receptive fields of each layer are of the same size, which constrains detection results of multiple scales pedestrians. In this paper, we propose a dynamic selection scheme to adaptive adjust receptive field size in multispectral pedestrian detection. Specifically, a network in network (NIN) is employed to combine visible and thermal information. Selective kernel networks (SKNets) which uses selective kernel unit with different kernel size are employed. To effectively fuse the feature representation in each layer, a build block is designed, in which different features are fused. Feature pyramid is employed to integrate feature information in each layer. We empirically show that our method outperforms existing 8 state-of-the-art methods on one multispectral dataset and 4 state-of-the-art methods on another multispectral dataset. Detailed analyses show that our proposed method can capture multispectral pedestrian detection of different scales, which confirms the effective of SKNets for adaptively resizing the receptive field sizes. In addition, our method can operate at 14 frames per second (fps) on a GPU.

NDGCN: Network in Network, Dilate Convolution and Graph Convolutional Networks Based Transportation Mode Recognition

Transportation mode recognition is a crucial task of Intelligent Transportation Systems (ITS) in smart city. Though many works have been investigated on transportation mode recognition in recent years, the accuracy and generality are still not able to meet the application requirements. In this paper, we propose a novel fusion framework for fine-grained transportation mode recognition, which consists of the Network in Network (NIN), Dilate Convolution and the Graph Convolutional Networks (GCN). In this framework, we first use NIN and Dilate Convolution to capture local and global features, respectively, and then introduce the graph convolutional network to learn the correlation of features. We construct a topological structure of the features based on the maximal information coefficient (MIC) criteria which is used to measure the similarity between two variables, and then obtain the adjacency matrix used for graph convolution. Extensive experimental results on the public Sussex-Huawei Locomotion-Transportation (SHL) dataset demonstrate the superiority of our proposed NDGCN to other state-of-the-art baselines with more than 22.3% higher accuracy.

New polyp image classification technique using transfer learning of network-in-network structure in endoscopic images

While colorectal cancer is known to occur in the gastrointestinal tract. It is the third most common form of cancer of 27 major types of cancer in South Korea and worldwide. Colorectal polyps are known to increase the potential of developing colorectal cancer. Detected polyps need to be resected to reduce the risk of developing cancer. This research improved the performance of polyp classification through the fine-tuning of Network-in-Network (NIN) after applying a pre-trained model of the ImageNet database. Random shuffling is performed 20 times on 1000 colonoscopy images. Each set of data are divided into 800 images of training data and 200 images of test data. An accuracy evaluation is performed on 200 images of test data in 20 experiments. Three compared methods were constructed from AlexNet by transferring the weights trained by three different state-of-the-art databases. A normal AlexNet based method without transfer learning was also compared. The accuracy of the proposed method was higher in statistical significance than the accuracy of four other state-of-the-art methods, and showed an 18.9% improvement over the normal AlexNet based method. The area under the curve was approximately 0.930 ± 0.020, and the recall rate was 0.929 ± 0.029. An automatic algorithm can assist endoscopists in identifying polyps that are adenomatous by considering a high recall rate and accuracy. This system can enable the timely resection of polyps at an early stage.

Vehicle Type Recognition Algorithm Based on Improved Network in Network

Vehicle type recognition algorithms are broadly used in intelligent transportation, but the accuracy of the algorithms cannot meet the requirements of production application. For the high efficiency of the multilayer perceptive layer of Network in Network (NIN), the nonlinear features of local receptive field images can be extracted. Global average pooling (GAP) can avoid the network from overfitting, and small convolution kernel can decrease the dimensionality of the feature map, as well as downregulate the number of model training parameters. On that basis, the residual error is adopted to build a novel NIN model by altering the size and layout of the original convolution kernel of NIN. The feasibility of the algorithm is verified based on the Stanford Cars dataset. By properly setting weights and learning rates, the accuracy of the NIN model for vehicle type recognition reaches 97.2%.

Fault Diagnosis of Conventional Circuit Breaker Contact System Based on Time–Frequency Analysis and Improved AlexNet

In order to eliminate the influence of parameter predefined caused by manual feature extraction, achieve fast feature extraction, and improve the recognition rate of fault diagnosis, a 2-D convolution neural network (CNN) method for fault diagnosis of conventional circuit breaker contact system is proposed. First, by introducing the data preprocessing method of continuous wavelet transform (CWT), the nonlinear and nonstationary original vibration signal is transformed into a time–frequency image to extract the transformed image features. Second, the convolutional layer module in the AlexNet model is combined with the network in network (NIN) module, and the global average pooling (GAP) layer is adopted to replace the fully connected (FC) layer, which realizes the improvement of the traditional AlexNet model. Then, an improved Adam optimization algorithm, namely, AMSGrad, is adopted to solve the problem that the Adam optimization algorithm may not converge or produce local optimization during model training. Finally, the preprocessed time–frequency image is taken as the input of the improved AlexNet model and through the supervised adjustment of network parameters, the fault diagnosis of the contact system for the conventional circuit breaker is realized accurately.

Diabetic Retinopathy Detection Using VGG-NIN a Deep Learning Architecture

Diabetic retinopathy (DR) is a disease that damages retinal blood vessels and leads to blindness. Usually, colored fundus shots are used to diagnose this irreversible disease. The manual analysis (by clinicians) of the mentioned images is monotonous and error-prone. Hence, various computer vision hands-on engineering techniques are applied to predict the occurrences of the DR and its stages automatically. However, these methods are computationally expensive and lack to extract highly nonlinear features and, hence, fail to classify DR’s different stages effectively. This paper focuses on classifying the DR’s different stages with the lowest possible learnable parameters to speed up the training and model convergence. The VGG16, spatial pyramid pooling layer (SPP) and network-in-network (NiN) are stacked to make a highly nonlinear scale-invariant deep model called the VGG-NiN model. The proposed VGG-NiN model can process a DR image at any scale due to the SPP layer’s virtue. Moreover, the stacking of NiN adds extra nonlinearity to the model and tends to better classification. The experimental results show that the proposed model performs better in terms of accuracy, computational resource utilization compared to state-of-the-art methods.

Deep network in network

The different CNN models use many layers that typically include a stack of linear convolution layers combined with pooling and normalization layers to extract the characteristics of the images. Unlike these models, and instead of using a linear filter for convolution, the network in network (NiN) model uses a multilayer perception (MLP), a nonlinear function, to replace the linear filter. This article presents a new deep network in network (DNIN) model based on the NiN structure, NiN drag a universal approximator, (MLP) with rectified linear unit (ReLU) to improve classification performance. The use of MLP leads to an increase in the density of the connection. This makes learning more difficult and time learning slower. In this article, instead of ReLU, we use the linear exponential unit (eLU) to solve the vanishing gradient problem that can occur when using ReLU and to speed up the learning process. In addition, a reduction in the convolution filters size by increasing the depth is used in order to reduce the number of parameters. Finally, a batch normalization layer is applied to reduce the saturation of the eLUs and the dropout layer is applied to avoid overfitting. The experimental results on the CIFAR-10 database show that the DNIN can reduce the complexity of implementation due to the reduction in the adjustable parameters. Also the reduction in the filters size shows an improvement in the recognition accuracy of the model.

Deep learned compact binary descriptor with a lightweight network-in-network architecture for visual description

Binary descriptors have been widely used for real-time image retrieval and correspondence matching. However, most of the learned descriptors are obtained using a large deep neural network (DNN) with several million parameters, and the learned binary codes are generally not invariant to many geometrical variances which is crucial for accurate correspondence matching. To address this problem, we proposed a new learning approach using a lightweight DNN architecture via a stack of multiple multilayer perceptrons based on the network in network (NIN) architecture, and a restricted Boltzmann machine (RBM). The latter is used for mapping the features to binary codes, and carry out the geometrically invariant correspondence matching task. Our experimental results on several benchmark datasets (e.g., Brown, Oxford, Paris, INRIA Holidays, RomePatches, HPatches, and CIFAR-10) show that the proposed approach produces the learned binary descriptor that outperforms other baseline self-supervised binary descriptors in terms of correspondence matching despite the smaller size of its DNN. Most importantly, the proposed approach does not freeze the features that are obtained while pre-training the NIN model. Instead, it fine-tunes the features while learning the features needed for binary mapping through the RBM. Additionally, its lightweight architecture makes it suitable for resource-constrained devices.

NF-RCNN: Heart localization and right ventricle wall motion abnormality detection in cardiac MRI

Convolutional neural networks (CNNs) are extensively used in cardiac image analysis. However, heart localization has become a prerequisite to these networks since it decreases the size of input images. Accordingly, recent CNNs benefit from deeper architectures in gaining abstract semantic information. In the present study, a deep learning-based method was developed for heart localization in cardiac MR images. Further, Network in Network (NIN) was used as the region proposal network (RPN) of the faster R-CNN, and then NIN Faster-RCNN (NF-RCNN) was proposed. NIN architecture is formed based on “MLPCONV” layer, a combination of convolutional network and multilayer perceptron (MLP). Therefore, it could deal with the complicated structures of MR images. Furthermore, two sets of cardiac MRI dataset were used to evaluate the network, and all the evaluation metrics indicated an absolute superiority of the proposed network over all related networks. In addition, FROC curve, precision-recall (PR) analysis, and mean localization error were employed to evaluate the proposed network. In brief, the results included an AUC value of 0.98 for FROC curve, a mean average precision of 0.96 for precision-recall curve, and a mean localization error of 6.17 mm. Moreover, a deep learning-based approach for the right ventricle wall motion analysis (WMA) was performed on the first dataset and the effect of the heart localization on this algorithm was studied. The results revealed that NF-RCNN increased the speed and decreased the required memory significantly.

Convolutional neural networks for multispectral pedestrian detection

In this paper, we present a pedestrian detection method by leveraging multispectral images which consist of color and thermal image information. Our method is based on the observation that a multispectral image enables us to overcome inherent limitations for pedestrian detection under challenging situations, e.g., insufficient illumination, small size pedestrian instances and occlusion. In order to detect pedestrian under such conditions, we apply deep convolutional neural networks (CNNs) for effectively combining color and thermal image information in multispectral images. We present a novel multispectral network that is built from the region-based fully convolutional networks (R-FCN) network model. A network-in-network (NIN) is employed to fuse these information across different modalities. Experimental results on KAIST benchmark demonstrate that our method surpasses the baseline method R-FCN and other proposed architectures.

Encrypted Network Traffic Classification Using Deep and Parallel Network-in-Network Models

Network traffic classification aims to recognize different application or traffic types by analyzing received data packets. This paper presents a neural network model with deep and parallel network-in-network (NIN) structures for classifying encrypted network traffic. Comparing with standard convolutional neural networks (CNN), NIN adopts a micro network after each convolution layer to enhance local modeling. Besides, NIN utilizes a global average pooling instead of traditional fully connected layers before final classification, which reduces the number of model parameters significantly. In our proposed method, deep NIN models with multiple MLP convolutional layers are built to map fixed-length packet vectors towards application or traffic labels. Furthermore, a parallel decision strategy of building two sub-networks to process packet header and packet body separately is designed considering that they may carry different kinds of clues for classification. The results of our experiments on the “ISCX VPN-nonVPN” encrypted traffic dataset show that NIN models can achieve a better balance between classification accuracy and model complexity than conventional CNNs. The parallel decision strategy can further improve the accuracy of using single NIN model for encrypted network traffic classification. Finally, the test set F1 scores of 0.983 and 0.985 are achieved for traffic characterization and application identification respectively.

Facial-image based Age Estimation Using Imbalanced Datasets

Facial image based human age estimation is of great application significance. The public-available facial image datasets used for age estimation suffer greatly from the uneven distribution of images of different age groups, which may lead to the low estimation accuracy of the under-sampled age categories and limit the usage of the age estimation in certain applications. We propose a three-stage probability adjustment based CNN algorithm to solve the imbalanced distribution problem of the dataset. In particular, we construct an ENIN neural network structure by applying the Network in Network (NIN) structure to the traditional convolution neural network (CNN) and use the probability vector adjustment to improve the classification accuracy of the under-sampled age categories. Then, we filter out the images with high possibility of being misclassified after the probability vector adjustment and reset their categories by comparing cosine similarity and retraining the ensembled ENIN classifier. We also introduce a population-age-distribution based accuracy metric Accuracy-P to estimate the performance of the age estimation algorithm in real-world applications. Our experimental results confirm that our algorithm can effectively improve the overall estimation accuracy by significantly improving the accuracy of the under-sampled age groups while maintaining satisfactory accuracy for the other age groups.