Max-pooling Operation Research Articles

An adaptive frame selection network with enhanced dilated convolution for video smoke recognition

Recognizing smoke from surveillance videos is crucial to achieve robust fire detection due to the rich temporal cues in video data. However, the slow-moving smoke makes it difficult to learn complete motion information of smoke. In addition, the widely-used max/average pooling operations in smoke recognition models may lead to the loss of spatial detail information. To solve the above issues, we propose an adaptive frame selection network (AFSNet) with enhanced dilated convolution for video smoke recognition. Specifically, first, to reduce information redundancy for learning discriminative feature representations, we propose an adaptive frame selection convolution to automatically select the most useful frames in the image sequence. Second, to utilize all the pixels for reducing the loss of detail information and carrying over global information, we propose an enhanced dilated convolution by considering the pixels with high responses as well as the local region to learn the contribution of each pixel to the output automatically. Finally, a feature extraction module without any average and max pooling operations is designed to learn multi-scale, context, and spatiotemporal information simultaneously. Experiments show that our model achieves the highest detection rate of 0.9673 and the lowest false alarm rate of 0.0316 on the SRSet dataset, and achieves the highest F1-score accuracy of 0.85, 0.86, and 0.91 on the three subsets of the RISE dataset, respectively.

Computer Aided COVID-19 Diagnosis in Pandemic Era Using CNN in Chest X-ray Images.

Early detection of abnormalities in chest X-rays is essential for COVID-19 diagnosis and analysis. It can be effective for controlling pandemic spread by contact tracing, as well as for effective treatment of COVID-19 infection. In the proposed work, we presented a deep hybrid learning-based framework for the detection of COVID-19 using chest X-ray images. We developed a novel computationally light and optimized deep Convolutional Neural Networks (CNNs) based framework for chest X-ray analysis. We proposed a new COV-Net to learn COVID-specific patterns from chest X-rays and employed several machine learning classifiers to enhance the discrimination power of the presented framework. Systematic exploitation of max-pooling operations facilitates the proposed COV-Net in learning the boundaries of infected patterns in chest X-rays and helps for multi-class classification of two diverse infection types along with normal images. The proposed framework has been evaluated on a publicly available benchmark dataset containing X-ray images of coronavirus-infected, pneumonia-infected, and normal patients. The empirical performance of the proposed method with developed COV-Net and support vector machine is compared with the state-of-the-art deep models which show that the proposed deep hybrid learning-based method achieves 96.69% recall, 96.72% precision, 96.73% accuracy, and 96.71% F-score. For multi-class classification and binary classification of COVID-19 and pneumonia, the proposed model achieved 99.21% recall, 99.22% precision, 99.21% F-score, and 99.23% accuracy.

Segmentation of shoulder muscle MRI using a new Region and Edge based Deep Auto-Encoder

Automatic segmentation of shoulder muscle MRI is challenging due to the high variation in muscle size, shape, texture, and spatial position of tears. Manual segmentation of tear and muscle portion is hard, time-consuming, and subjective to pathological expertise. This work proposes a new Region and Edge-based Deep Auto-Encoder (RE-DAE) for shoulder muscle MRI segmentation. The proposed RE-DAE harmoniously employs average and max-pooling operations in the Convolutional encoder and decoder blocks. The DAE’s region and edge-based segmentation encourage the network to extract homogenous and anatomical information, respectively. These two concepts, systematically combined in a DAE, generate a discriminative and sparse hybrid feature space (exploiting both region homogeneity and boundaries). Moreover, the concept of static attention is exploited in the proposed RE-DAE that helps effectively learn the tear region. The performances of the proposed RE-DAE architectures have been tested using a 3D MRI shoulder muscle dataset using the hold-out cross-validation technique. The MRI data has been collected from the Korea University Anam Hospital, Seoul, South Korea. Experimental comparisons have been conducted by employing innovative custom-made and existing pre-trained CNN architectures using transfer learning and fine-tuning. Objective evaluation on the muscle datasets using the proposed SA-RE-DAE showed a dice similarity of 85.58% and 87.07%, an accuracy of 81.57% and 95.58% for tear and muscle regions, respectively. The high visual quality and the objective result suggest that the proposed SA-RE-DAE can correctly segment tear and muscle regions in shoulder muscle MRI for better clinical decisions.

Attention in Attention: Modeling Context Correlation for Efficient Video Classification

Attention mechanisms have significantly boosted the performance of video classification neural networks thanks to the utilization of perspective contexts. However, the current research on video attention generally focuses on adopting a specific aspect of contexts (e.g., channel, spatial/temporal, or global context) to refine the features and neglects their underlying correlation when computing attentions. This leads to incomplete context utilization and hence bears the weakness of limited performance improvement. To tackle the problem, this paper proposes an efficient attention-in-attention (AIA) method for element-wise feature refinement, which investigates the feasibility of inserting the channel context into the spatio-temporal attention learning module, referred to as CinST, and also its reverse variant, referred to as STinC. Specifically, we instantiate the video feature contexts as dynamics aggregated along a specific axis with global average and max pooling operations. The workflow of an AIA module is that the first attention block uses one kind of context information to guide the gating weights calculation of the second attention that targets at the other context. Moreover, all the computational operations in attention units act on the pooled dimension, which results in quite few computational cost increase (< 0.02%). To verify our method, we densely integrate it into two classical video network backbones and conduct extensive experiments on several standard video classification benchmarks. The source code of our AIA is available at <uri xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">https://github.com/haoyanbin918/Attention-in-Attention</uri> .

Bidirectional Long Short-Term Memory with Temporal Dense Sampling for human action recognition

Long Short-Term Memory networks are making significant inroads into improving time series applications, including human action recognition. In a human action video, the spatial and temporal streams carry distinctive yet prominent information, hence many researchers turn to spatio-temporal models for human action recognition. A spatio-temporal model integrates the temporal network (e.g. Long Short-Term Memory) and spatial network (e.g. Convolutional Neural Networks). There are few challenges in the existing human action recognition: (1) the uni-directional modeling of Long Short-Term Memory making it unable to preserve the information from the future, (2) the sparse sampling strategy tends to lose prominent information when performing dimension reduction on the input of Long Short-Term Memory, and (3) the fusion strategy for consolidating the temporal network and spatial network. In view of this, we propose a Bidirectional Long Short-Term Memory with Temporal Dense Sampling and Fusion Network method to address the above-mentioned challenges. The Temporal Dense Sampling partitions the human action video into segments and then performs maxpooling operation along the temporal axis in each segment. A multi-stream bidirectional Long Short-Term Memory network is adopted to encode the long-term spatial and temporal dependencies in both forward and backward directions. Instead of assigning fixed weights to the spatial network and temporal network, we propose a fusion network where a fully-connected layer is trained to adaptively assign the weights for the networks. The empirical results demonstrate that the proposed Bidirectional Long Short-Term Memory with Temporal Dense Sampling and Fusion Network method outshines the state-of-the-art methods with an accuracy of 94.78% on UCF101 dataset and 70.72% on HMDB51 dataset.

Intelligent Diagnosis of Fish Behavior Using Deep Learning Method

Scientific methods are used to monitor fish growth and behavior and reduce the loss caused by stress and other circumstances. Conventional techniques are time-consuming, labor-intensive, and prone to accidents. Deep learning (DL) technology is rapidly gaining popularity in various fields, including aquaculture. Moving towards smart fish farming necessitates the precise and accurate identification of fish biodiversity. Observing fish behavior in real time is imperative to make better feeding decisions. The proposed study consists of an efficient end-to-end convolutional neural network (CNN) classifying fish behavior into the normal and starvation categories. The performance of the CNN is evaluated by varying the number of fully connected (FC) layers with or without applying max-pooling operation. The accuracy of the detection algorithm is increased by 10% by incorporating three FC layers and max pooling operation. The results demonstrated that the shallow architecture of the CNN model, which employs a max-pooling function with more FC layers, exhibits promising performance and achieves 98% accuracy. The presented system is a novel step in laying the foundation for an automated behavior identification system in modern fish farming.

Discriminative Convolutional Sparse Coding of ECG Signals for Automated Recognition of Cardiac Arrhythmias

Electrocardiogram (ECG) is a common and powerful tool for studying heart function and diagnosing several abnormal arrhythmias. In this paper, we present a novel classification model that combines the discriminative convolutional sparse coding (DCSC) framework with the linear support vector machine (LSVM) classification strategy. In the training phase, most existing convolutional sparse coding frameworks are unsupervised in the sense that label information is ignored in the convolutional filter training stage. In this work, we explicitly incorporate a label consistency constraint called “discriminative sparse-code error” into the objective function to learn discriminative dictionary filters for sparse coding. The learned dictionary filters encourage signals from the same class to have similar sparse codes, and signals from different classes to have dissimilar sparse codes. To reduce the computational complexity, we propose to perform a max-pooling operation on the sparse coefficients. Using LSVM as a classifier, we examine the performance of the proposed classification system on the MIT-BIH arrhythmia database in accordance with the AAMI EC57 standard. The experimental results show that the proposed DCSC + LSVM algorithm can obtain 99.32% classification accuracy for cardiac arrhythmia recognition.

ELEKTRA: ELEKTRokardiomatrix application to biometric identification with convolutional neural networks

Biometric systems are an uprising technique of identification in today’s world. Many different systems have been used in everyone’s daily life in the past years, such as fingerprint, face scan, and others. We propose a new identification method using Elektrokardiogramms (EKGs) converted into a heatmap of a set of aligned R-peaks (heartbeats), forming a matrix called an Elektrokardiomatrix (EKM). We can build a one-against-many identification system using a Convolutional Neural Network (CNN). We have tested our proposal with one main database (the Normal Sinus Rhythm Database (NSRDB)) and two other databases, which are the MIT-BIH Arrhythmia Database (MIT-BIHDB) and the Physikalisch-Technische Bundesanstalt (PTB) Database. With the NSRDB, we have achieved an accuracy of 99.53% and offered a False Acceptance Rate (FAR) of 0.02% and a False Rejection Rate (FRR) of 0.05%. Very similar results were also obtained with the MIT-BIH and PTB databases. We have performed in-depth experimentation to test the efficiency and feasibility of our novel biometric solution. It is remarkable that with a simple CNN, which has only one convolutional layer, a max-pooling operation, and some regularisation, we can identify users with very high performance and low error rates. Consequently, our model does not need very complex architectures to offer high-performance metrics.

An Application of Pixel Interval Down-Sampling (PID) for Dense Tiny Microorganism Counting on Environmental Microorganism Images

This paper proposes a novel pixel interval down-sampling network (PID-Net) for dense tiny object (yeast cells) counting tasks with higher accuracy. The PID-Net is an end-to-end convolutional neural network (CNN) model with an encoder–decoder architecture. The pixel interval down-sampling operations are concatenated with max-pooling operations to combine the sparse and dense features. This addresses the limitation of contour conglutination of dense objects while counting. The evaluation was conducted using classical segmentation metrics (the Dice, Jaccard and Hausdorff distance) as well as counting metrics. The experimental results show that the proposed PID-Net had the best performance and potential for dense tiny object counting tasks, which achieved 96.97% counting accuracy on the dataset with 2448 yeast cell images. By comparing with the state-of-the-art approaches, such as Attention U-Net, Swin U-Net and Trans U-Net, the proposed PID-Net can segment dense tiny objects with clearer boundaries and fewer incorrect debris, which shows the great potential of PID-Net in the task of accurate counting.

Architectural and mathematical fundamentals of improvement neural networks for classification of images

The article proposes ways to solve the problem of structural synthesis of neural networks architectures; based on non-traditional approaches to their analytical formalization and application of new ones data processing operations. The example of the CIFAR10 dataset proves the possibility of improving the accuracy of the solution image classification tasks within an alternative architecture with expandable input and narrowing structures of trained neural networks. The results of the application of input expanding image taken with Resizing and Conv2DTranspose layers at the input trained neural networks indicate the effectiveness of solving such classification problems by example CIFAR10 dataset. Achieved on the basis of the Xception neural network, the average classification accuracy of 10 classes CIFAR10 images are 97.3%. The side effect of pre-scaling images is leveling the accuracy of the classification of different classes, which allows us to consider such a change in size as an option data augmentation in the dataset. Tensor-matrix methods have been introduced for further development of this approach formalization of the description of neural networks on the basis of the penetrating end product of matrices and its block ones modifications. On this basis, a number of new convolution and maxpooling operations are proposed, as well as combinations on input of the narrowing segment not only symmetrically enlarged image, but also its variants obtained on the basis of a generalized penetrating product. It is about drawing images expanded by rows of pixels (horizontally) and columns (vertically), as well as a combination of several different options symmetrical image extensions on the principle of constructing a pyramidal segment of the PSPNet neural network.

Improved Classification Model using CNN for Detection of Alzheimer’s Disease

Alzheimer’s Disease (AD) is commonly called a neurodegenerative disorder and it is a common form of dementia. There is no permanent cure for this brain disease hence the early diagnosis of such disease using medical imaging system is highly significant. Machine learning models play a vital role in the detection of AD. Since most of the conventional machine learning models find it difficult to detect the essential features to classify the disease, an advanced deep learning framework called Convolutional Neural Network (CNN) is used in this study to detect essential features automatically and classify the disease. The building components of the proposed CNN-based classification method include convolution layer, batch normalization process, ReLU, and Max-pooling operation. The main objective of this CNN-based classification method is to predict whether the patient is suffering from Alzheimer's disease through the analysis of brain MRI. The proposed methodology implemented is identical to a classification-based system that undergoes training, evaluation, and testing process. Finally, the softmax layer is applied for classification, and the Adam optimization technique is applied for reducing the loss, and by applying Adam quicker convergence can be achieved. The proposed improved CNN classification method achieves an accuracy of 97.8%.

A wavelet convolutional capsule network with modified super resolution generative adversarial network for fault diagnosis and classification

The study of fault diagnosis and classification has gained tremendous attention in various aspects of modern industry. However, the performance of traditional fault diagnosis technique solely depends on handcrafted features based on expert knowledge which is difficult to pre-design and has failed in several applications. Deep learning (DL) has achieved remarkable performance in hierarchical feature extraction and learning distinctive feature of dataset from related distribution. However, the challenge associated with DL models is that max-pooling operation usually leads to loss of spatial details during high-level feature extraction. Another concern is the low quality characteristics of 2D time-frequency image which is mostly caused by the presence of noise and poor resolution. This paper proposes a modified wavelet convolutional capsule network with modified enhanced super resolution generative adversarial network plus for fault diagnosis and classification. It uses continuous wavelet transform to convert raw data signals to 2D time-frequency images and applies super resolution generative adversarial technique to enhance the quality of the time-frequency images and finally, the convolutional capsule network learns the extracted high-level features without loss of spatial details for the diagnosis and classification of faults. We validated our proposed model on the famous motor bearing dataset from the Case Western Reserve University. The experimental results show that our proposed fault diagnostic model obtains higher diagnosis accuracy of 99.84% outweighing most traditional deep learning models including state-of-the-art methods.

Detecting retinal vasculature as a key biomarker for deep Learning-based intelligent screening and analysis of diabetic and hypertensive retinopathy

Retinal vessels are considered important biomarkers for the detection of retinal diseases, like diabetic retinopathy (caused by diabetes) and hypertensive retinopathy (caused by hypertension). The manual finding from this retinal vasculature is time-consuming and costly. The image quality of the fundus image directly affects the accurate segmentation of these vessels in the automatic methods. With such inferior quality images, deep-learning-based methods are better for dealing with segmentation. Conventional deep-learning-based vessel segmentation methods deal the segmentation task with deeper convolutional neural networks and many trainable parameters. Minor changes in the retinal vasculature, such as those that result from the creation of new smaller vessels, is crucial for the keen analysis of diseases (e.g., diabetic retinopathy). The small vessels are crucial to segment, owing to continuous max-pooling operations and deeper networks. We herein present a pool-less residual segmentation network that is capable of segmenting even smaller vessels using a shallower network with a low number of trainable parameters. Our proposed pool-less residual segmentation network (PLRS-Net) is a vessel segmentation network that provides the pooling effect with strided convolution for better segmentation sensitivity. The final PLRS-Net is an advanced form of a pool-less segmentation network (PLS-Net) wherein semantic segmentation was performed with a few layers, and the residual connection fulfilled the feature enhancement strategy to construct PLRS-Net. PLS-Net and PLRS-Net are two separate networks that can perform vessel segmentation without prior preprocessing and postprocessing.To evaluate our proposed method, the experiments include three publicly available datasets: Digital retinal images for vessel extraction (DRIVE), child heart health study in England database (CHASE-DB1), and structured analysis of retina (STARE). The results demonstrate that our proposed method provides a high segmentation performance, achieving an average sensitivity (Sen) of 82.69, specificity (Spe) of 98.17, accuracy (Acc) 96.82, and area under the curve (AUC) of 98.35 for the DRIVE dataset, Sen of 83.01, Spe of 98.39, Acc of 97.31, and AUC of 98.63 for the CHASE-DB1 dataset, and a Sen of 86.35, Spe of 98.03, Acc of 97.15, and AUC of 98.99 for the STARE dataset. These accuracies show exceptional segmentation performance of the proposed method compared to state-of-the-art approaches for automatic vessel detection for diagnosis purposes.

A systematic extraction of glacial lakes for satellite imagery using deep learning based technique

For outburst susceptibility assessment of glacial lakes, the accurate extraction of glacial lakes region from satellite image is essential. Several techniques are introduced to extract glacial lakes over the years, such as based on image pixel or different thresholds and object-based statistics. These methods require manual optimization parameters for accurate extraction of the glacial lake’s region. Hence, in this work, the deep learning (DL) models are utilized to segment the water and non-water bodies from satellite imagery. DL techniques are successful in many research areas for classification problems, but these techniques are relatively new for the extraction of glacial lakes. The main challenges in implementing DL techniques for the extraction of the glacial lakes are (i) preserving the semantic features of the image without loss during max-pooling operation, (ii) huge dataset requirement for proper training. To overcome the problems mentioned above, a customized architecture of the U-Net named as glacial lakes U-Net (GLU-Net) with various encoders is proposed in this work for effective segmentation of the glacial lakes. The proposed method for segmentation is also compared with the existing DL techniques in this work. The DL models are evaluated on Imja, Chandra, and Bhaga glacier regions datasets collected from Landsat 8 satellite imagery. The average sensitivity (Sen), specificity (Spe), and dice coefficient (Dice) for GLU-Net are 92.23%, 99.81%, and 93.46%. These are the three measures for an average of Imja, Chandra and Bhaga glacier regions. The qualitative and quantitative performance analysis shows the significant improvement of the proposed technique compared to other DL techniques for extraction of glacial lake’s region.

Classification and Region Analysis of COVID-19 Infection Using Lung CT Images and Deep Convolutional Neural Networks

COVID-19 is a global health problem. Consequently, early detection and analysis of the infection patterns are crucial for controlling infection spread as well as devising a treatment plan. This work proposes a two-stage deep Convolutional Neural Networks (CNNs) based framework for delineation of COVID-19 infected regions in Lung CT images. In the first stage, initially, COVID-19 specific CT image features are enhanced using a two-level discrete wavelet transformation. These enhanced CT images are then classified using the proposed custom-made deep CoV-CTNet. In the second stage, the CT images classified as infectious images are provided to the segmentation models for the identification and analysis of COVID-19 infectious regions. In this regard, we propose a novel semantic segmentation model CoV-RASeg, which systematically uses average and max pooling operations in the encoder and decoder blocks. This systematic utilization of max and average pooling operations helps the proposed CoV-RASeg in simultaneously learning both the boundaries and region homogeneity. Moreover, the idea of attention is incorporated to deal with mildly infected regions. The proposed two-stage framework is evaluated on a standard Lung CT image dataset, and its performance is compared with the existing deep CNN models. The performance of the proposed CoV-CTNet is evaluated using Mathew Correlation Coefficient (MCC) measure (0.98) and that of proposed CoV-RASeg using Dice Similarity (DS) score (0.95). The promising results on an unseen test set suggest that the proposed framework has the potential to help the radiologists in the identification and analysis of COVID-19 infected regions.

Acoustic Source Localization Using a Geometrically Sampled Grid SRP-PHAT Algorithm With Max-Pooling Operation

The steered response power phase transform (SRP-PHAT) is a well-known algorithm for acoustic source localization using microphone arrays. It consists in the computation of the generalized cross-correlation (GCC) between each microphone pair, and in the coherent summation of the GCC values in the grid search space. Several improvements based on the volumetric grid have been proposed in order to achieve spatial resolution scalability and to reduce the computational cost by using a coarser grid. In general, the problem of the volumetric based methods is that the noise and the reverberation are projected into the search space since all GCC information is used to build the acoustic map. It is hence proposed a volumetric grid SRP-PHAT algorithm based on the geometrically sampled grid (GSG) that incorporates a max-pooling (MP) operation in the volume accumulation of the GCC values in order to improve the localization performance. The MP is the solution of a minimization-maximization problem that aims at minimizing the deleterious effect of noise and reverberation and at maximizing the accuracy of the GCC values related to the target sound source. Simulations and real-world experiments demonstrate the efficiency of the proposed SRP-GSG-MP algorithm in adverse conditions.

Cascaded Multiscale Structure With Self-Smoothing Atrous Convolution for Semantic Segmentation

Convolutional neural networks (CNNs) have attracted great attention in the semantic segmentation of very-high-resolution (VHR) images of urban areas. However, large-scale variation of objects in the urban areas often makes it difficult to achieve good segmentation accuracy. Atrous convolution and atrous spatial pyramid pooling composed of atrous convolution can alleviate this problem by exploring multiscale contextual information. Unfortunately, atrous convolution causes gridding artifacts, where actual receptive fields are separated unit sets and fail to cover all the receptive fields. To address this problem, in this article, we first propose a self-smoothing atrous convolution (SS-AConv) that intrinsically improves atrous convolution, unlike existing methods. SS-AConv enhances sampling rates with low computational costs by adding several key parameters in the spatial dimension of its filter. Then, in the backbone, the Xception network, all max pooling operations are replaced with SS-AConvs for a large and effective receptive field. Moreover, to extract diverse features at multiple scales, we propose the SS-AConv cascaded multiscale structure (SCMS) by integrating SS-AConvs with different rates and the residual correction scheme (RCS) into a cascaded spatial pyramid. Finally, to extract diverse features at dense multiple scales, the SS-AConv convolutional network (SS-ACNet) is constructed by integrating SCMS into both the encoder and decoder layers of the modified Xception network. Extensive experimental results show that SS-ACNet outperforms some state-of-the-art methods on four open challenge datasets: the ISPRS Vaihingen and Potsdam, Cityscapes, and PASCAL VOC 2012.

Drone Navigation Using Region and Edge Exploitation-Based Deep CNN

Drones are unmanned aerial vehicles (UAV) utilized for a broad range of functions, including delivery, aerial surveillance, traffic monitoring, architecture monitoring, and even War-field. Drones confront significant obstacles while navigating independently in complex and highly dynamic environments. Moreover, the targeted objects within a dynamic environment have irregular morphology, occlusion, and minor contrast variation with the background. In this regard, a novel deep Convolutional Neural Network(CNN) based data-driven strategy is proposed for drone navigation in the complex and dynamic environment. The proposed Drone Split-Transform-and-Merge Region-and-Edge (Drone-STM-RENet) CNN is comprised of convolutional blocks where each block methodically implements region and edge operations to preserve a diverse set of targeted properties at multi-levels, especially in the congested environment. In each block, the systematic implementation of the average and max-pooling operations can deal with the region homogeneity and edge properties. Additionally, these convolutional blocks are merged at a multi-level to learn texture variation that efficiently discriminates the target from the background and helps obstacle avoidance. Finally, the Drone-STM-RENet generates steering angle and collision probability for each input image to control the drone moving while avoiding hindrances and allowing the UAV to spot risky situations and respond quickly, respectively. The proposed Drone-STM-RENet has been validated on two urban cars and bicycles datasets: udacity and collision-sequence, and achieved considerable performance in terms of explained variance (0.99), recall (95.47%), accuracy (96.26%), and F-score (91.95%). The promising performance of Drone-STM-RENet on urban road datasets suggests that the proposed model is generalizable and can be deployed for real-time autonomous drones navigation and real-world flights.

Fusion of multi-scale bag of deep visual words features of chest X-ray images to detect COVID-19 infection

Chest X-ray (CXR) images have been one of the important diagnosis tools used in the COVID-19 disease diagnosis. Deep learning (DL)-based methods have been used heavily to analyze these images. Compared to other DL-based methods, the bag of deep visual words-based method (BoDVW) proposed recently is shown to be a prominent representation of CXR images for their better discriminability. However, single-scale BoDVW features are insufficient to capture the detailed semantic information of the infected regions in the lungs as the resolution of such images varies in real application. In this paper, we propose a new multi-scale bag of deep visual words (MBoDVW) features, which exploits three different scales of the 4th pooling layer’s output feature map achieved from VGG-16 model. For MBoDVW-based features, we perform the Convolution with Max pooling operation over the 4th pooling layer using three different kernels: 1 times 1, 2 times 2, and 3 times 3. We evaluate our proposed features with the Support Vector Machine (SVM) classification algorithm on four CXR public datasets (CD1, CD2, CD3, and CD4) with over 5000 CXR images. Experimental results show that our method produces stable and prominent classification accuracy (84.37%, 88.88%, 90.29%, and 83.65% on CD1, CD2, CD3, and CD4, respectively).

Convolutional Neural Network Method for Classification of Syllables in Javanese Script

Javanese script is one of the languages which are a typical Javanese culture. Javanese script is seen in its use in writing the name of a particular agency or location that has historical and tourism value. The use of Javanese script in public places makes the existence of this script seen by many people, not only by the Javanese people. Some of them have difficulty recognizing the Javanese characters they encounter. One method of pattern recognition and image processing is Convolutional Neural Network (CNN). CNN is a method that uses convolution operations in performing feature extraction on images as a basis for classification. The process consists of initial data processing, classification, and syllable formation. The classification consists of 48 classes covering Javanese script types, namely basic letters (Carakan) and voice-modifying scripts (Sandhangan). It is tested with multi-class confusion matrix scenarios to determine the accuracy, precision, and recall of the built CNN model. The CNN architecture consists of three convolution layers with max-pooling operations. The training configuration includes a learning rate of 0.0001, and the number of filters for each convolution layer is 32, 64, and 128 filters. The dropout value used is 0.5, and the number of neurons in the fully-connected layer is 1,024 neurons. The average performance value of accuracy reached 87.65%, the average precision value was 88.01%, and the average recall value was 87.70%.