Deep Convolutional Neural Network Classifier Research Articles

Chicken-Moth Search Optimization-based Deep Convolutional Neural Network for Image Steganography

Image steganography is considered as one of the promising and popular techniques utilized to maintain the confidentiality of the secret message that is embedded in an image. Even though there are various techniques available in the previous works, an approach providing better results is still the challenge. Therefore, an effective pixel prediction based on image stegonography is developed, which employs error dependent Deep Convolutional Neural Network (DCNN) classifier for pixel identification. Here, the best pixels are identified from the medical image based on DCNN classifier using pixel features, like texture, wavelet energy, Gabor, scattering features, and so on. The DCNN is optimally trained using Chicken-Moth search optimization (CMSO). The CMSO is designed by integrating Chicken Swarm Optimization (CSO) and Moth Search Optimization (MSO) algorithm based on limited error. Subsequently, the Tetrolet transform is fed to the predicted pixel for the embedding process. At last, the inverse tetrolet transform is used for extracting the secret message from an embedded image. The experimentation is carried out using BRATS dataset, and the performance of image stegonography based on CMSO-DCNN+tetrolet is evaluated based on correlation coefficient, Structural Similarity Index, and Peak Signal to Noise Ratio, which attained 0.85, 46.981dB, and 0.6388, for the image with noise.

Introduction to the Special Issue on Intelligence on Scalable computing for Recent Applications

Introduction to the Special Issue on Intelligence on Scalable computing for Recent Applications

An abnormality detection of retinal fundus images by deep convolutional neural networks

Identification of retinal diseases is a test for the ophthalmologists as the anomalies are just unmistakable at the beginning period. Early detection of these diseases can avoid lasting vision misfortune. Dealing with a lot of retinal images and location of variations from the norm because of these infections is difficult just as tedious. In this work, the deep learning algorithm has proposed to check the abnormality condition of retina with the help of retinal fundus images. In deep leaning a training set is produced with features of variations from the norm present in the retinal images and the infection the retina is experiencing. The deep Convolutional Neural Network (CNN) classifier predicts the infection for every retinal images in the wake of social event the learning from training the set. The rightness of desire is resolved to evaluate the viability of the classifier. The proposed technique was executed in MATLAB and assessed both normal and abnormal diabetic retinopathy retinal images of IDRID, ROC, and local datasets. The proposed technique has gotten better execution measurements, for example, sensitivity of 98.2%, Specificity of 98.45%, accuracy of 98.56% and average area under receiver operating characteristics of 0.9 when contrasted with different conditions of the workmanship strategies.

Handwritten Kannada numerals recognition using deep learning convolution neural network (DCNN) classifier

In pattern recognition, identifying Kannada handwritten numerals are a complex knot. This paper portrays an avenue that pikes us to attain a most potent Kannada Numerals recognition process. In this, handwritten Kannada characters are captivated in document fashion and are subjected to Pre-processing and attribute extraction processes. Pre-processing entails steps like noise removal, binarization, normalization, skew amendment, and thinning. Features are extricated by exploiting strategies like Drift Length Count, Direction related progression code, DWT and Curvelet Transfiguration Wrapping. For an impressive classification process deep convolution neural network classifier is preferred. Isolation accuracy of Kannada numeral aimed here will outsource 96% of accuracy.

Effective Video Saliency Mapping for Object Detection using Deep Learning Neural Networks

In this paper, we propose a new novel DNN-based video saliency prediction method.The main aim of this project is to separating foreground and background images and also finding the motion of the objects.The methodogy used in this project is to seperating the foreground images using background modelling techniques(Pixel searching algorithm).The another technique is Homography based search algorithm for block based motion estimation is used to finding the movement of the objects.Saliency object detection is also used to highlighting the particular region in the images.In our base paper[1],context subtraction algorithm is introduced.But in this paper,Convolutional neural networks is used to classifying the foreground objects from background images.We propose an deep convolutional neural network (CNN) Classifier is used to predict intra-frame salience by exploring object and object movement details .The CNN classifier is also used to separating the new objects from background. We also find from our database that a temporary correlation of human attention exists, with a smooth transition of saliency across video frames.Therefore, in our DNN-based approach we create a two-layer convolutionary long-term memory (2C-LSTM) network, using the extracted CNN features as the data.The fuzzy clustering is used to store the information as the processed images in this project.This project can be implemented in real time applications such as defence,forest and miltrary area.Saliency Mapping is also used to reduce the Zipper effect,it mainly occurs in edges.It gives High resolution too. Finally, the proposed system is developed using matlab simulation.The PSNR value is attained above 95%.

Deep learning-based roadway crack classification using laser-scanned range images: A comparative study on hyperparameter selection

In recent years, deep learning-based crack detection methods have been widely explored and applied due to their high versatility and adaptability. In civil engineering applications, recent research on crack detection through deep convolutional neural network (DCNN) includes road pavement crack detection, bridge inspection, defects detection in shield tunnel lining, etc. Despite the increasing popularity of DCNN on crack detection, many challenges have yet to be properly addressed. For crack detection using three-dimensional (3D) range (i.e., elevation) images, disturbances such as surface variation can negatively affect the detection performance. Besides, some typical non-crack patterns such as grooves can be easily misidentified as cracks, i.e., false positives. Another issue lies in the selection of hyperparameters related with the design of a DCNN architecture. For example, the hyperparameters which are related with network structure (e.g., kernel size, network depth and width) and training (e.g., mini-batch size and learning rate) can impact the network performance to a significant extent. Therefore, they need to be properly determined for optimal performance. However, for deep learning-based roadway crack classification using laser-scanned range images, a comprehensive discussion on the hyperparameter selection/tuning has not been thoroughly performed. This study develops a hyperparameter selection process involving a series of experiments on laser-scanned range images with high diversities, investigating the optimal joint hyperparameter configuration on network structure and training for DCNN-based roadway crack classification. In a comparative study, 36 DCNN architectures with varying layouts are developed for crack classification. These architecture candidates differ in kernel sizes (e.g., 3 × 3, 7 × 7, and 11 × 11), network depths (from 5 to 8 weight layers), and widths (from 16 to 96 kernels in each convolutional layer). The 7-layer DCNN with constant 7 × 7 kernels and increasing network widths yields the highest classification performance among the proposed 36 DCNN classifiers, which may be because it can best reflect the complexity of the acquired laser-scanned roadway range images. Once the optimal architecture layout is determined, further discussion on the selection of min-batch sizes, learning rates, dropout factor and leaky rectified linear unit (LReLU) factor is performed. Experimental results show the optimal architecture with associated training configuration can achieve consistent and accurate performance, under the contamination of surface variations and grooved patterns in laser-scanned range images. Discussion on the hyperparameter selection can provide insights for the development of DCNN in similar applications using laser-scanned range images.

Computer-aided diagnosis for burnt skin images using deep convolutional neural network

Numerous patients died every year due to the leading causes of deaths all over the world and burn injuries are one of them. Burn injury cases are most viewed in low and middle-income countries (LMIC). Researchers show great interest to classify the burn into different depths through digital means. In Pakistan, at provisional level, it’s really a significant issue to categorize the burn and its depths due to the non-availability of expert doctors and surgeons; hence the decision for the correct first treatment can't be made, so this may cause a serious issue later on. The main objectives of this research work are to segment the burn wounds and classification of burn depths into 1st, 2nd and 3rd degrees respectively. A real-time dataset of burnt patients has been collected from the burn unit of Allied Hospital Faisalabad, Pakistan. The dataset used for this research task contains 450 images of all the three levels of burn depths. Segmentation of the burnt area was done by the use of Otsu's method of thresholding and feature vector was obtained through the use of statistical methods. We have used the Deep Convolutional Neural Network (DCNN) to estimate the burn depths. The network was trained by 65 percent of the images and the remaining 35 percent images were used for testing the accuracy of the classifier. The maximum average accuracy obtained by using the Deep Convolutional Neural Network (DCNN) classifier is reported round about 79.4% and these results are the best if we compare them with previous results. From the obtained results of this research work, non-expert doctors will be able to apply the correct first treatment for the quality evaluation of burn depths.

Deep Convolutional Neural Networks models for the binary morphological classification of SDSS-galaxies

We present the deep learning approach for the determination of morphological types of galaxies. We demonstrate the method's performance with the redshift-limited (z < 0.1) training sample of 6 163 galaxies from the SDSS DR9. We exploited the deep convolutional neural network classifiers such as InceptionV3, DenseNet121, and MobileNetV2 to process images of SDSS-galaxies (100x100 pixels, 25 arcsec in each axis in size) using g, r, i filters as R - G - B channels to create images. We provided the data augmentation (horizontal and vertical flips, random shifts on ±10 pixels, and rotations) randomly applied to the set of images during learning, which helped increase the classifier's generalization ability. Also, two different loss functions, MAE and Lovasz-Softmax, were applied to each classifier. The target sample galaxies were classified into two morphological types (late and early) trained on the images of galaxies from the sample. It turned out that the deep convolutional neural networks InceptionV3 and DenseNet121 with MAE-loss function show the best result attaining 93.3% accuracy.

Method for Classifying a Noisy Raman Spectrum Based on a Wavelet Transform and a Deep Neural Network

Because it is relatively difficult in practice to classify the Raman spectrum under baseline noise and additive white Gaussian noise environments, this paper proposes a new framework based on a wavelet transform and deep neural network for identification of noisy Raman spectra. The framework consists of two main engines. Wavelet transform is proposed as the framework front end for transforming the 1-D noise Raman spectrum to two-dimensional data. The two-dimensional data are fed to the framework back end, which is a classifier. The optimum classifier is chosen by implementing several traditional machine learning (ML) and deep learning (DL) algorithms, and we investigate their classification accuracy and robustness performances. The four chosen MLs are naive Bayes (NB), a support vector machine (SVM), a random forest (RF) and a k-nearest neighbor (KNN), and a deep convolution neural network (DCNN) was chosen as a DL classifier. Noise-free, Gaussian noise, baseline noise, and mixed-noise Raman spectra were applied to train and validate the ML and DCNN models. The optimum back-end classifier was obtained by testing the ML and DCNN models with several noisy Raman spectra (10-30 dB noise power). Based on the simulation, the accuracy of the DCNN classifier is 9% higher than that of the NB classifier, 3.5% higher than the RF classifier, 1% higher than the KNN classifier, and 0.5% higher than the SVM classifier. In terms of robustness to mixed noise scenarios, the framework with the DCNN back end showed superior performance compared with the other ML back ends. The DCNN back end achieved 90% accuracy at 3 dB SNR, while the NB, SVM, RF, and K-NN back ends required 27 dB, 22 dB, 27 dB, and 23 dB SNR, respectively. In addition, in the low-noise test dataset, the F-measure score of the DCNN back end exceeded 99.1%, and the F-measure scores of the other ML engines were below 98.7%.

Monarch-EWA: Monarch-Earthworm-Based Secure Routing Protocol in IoT

AbstractRouting in the Internet of Things (IoT) renders the protection against various network attacks as any attacker intrudes the routing mechanism for establishing the destructive mechanisms against the network, which insists the essentiality of the security protocols in IoT. Thus, the paper proposes a secure protocol based on an optimization algorithm, Monarch-Earthworm Algorithm (Monarch-EWA), which is the modification of the Monarch Butterfly algorithm using the Earthworm Optimization Algorithm (EWA) in order to render effective security to the network. Initially, the effective nodes are selected using the Deep Convolutional Neural Network (deep CNN) classifier based on the factors, trust and energy of the node, and stochastic gradient descent algorithm trains the deep CNN classifier. The secure nodes are involved in routing for which the secure multipath is chosen optimally using the proposed Monarch-EWA, which chooses the secure multipath based on the factors, energy and trust. The analysis of the proposed method in the presence of attacks, such as black hole, message replicate and distributed denial of service, reveals that the proposed method outperformed the existing methods. The proposed Monarch-EWA protocol acquired the maximal energy, throughput and detection rate of 0.2268 J, 48.2759% and 82.6231%, respectively, with the minimal delay of 0.0959 ms.

(DCNN) Deep Convolution Neural Network Classifier and (EW-CSA) Earth Worm-Crow Search Algorithm for Lane Detection

Every year in India, most of the car accidents are occurs and affects on number of lives. Most of the road accidents are occurs due to driver’s inattention and fatigue. Drivers require to focus on different circumstances, together with vehicle speed and path, the separation between vehicles, passing vehicles, and potential risky or uncommon events ahead. Also the accident occurs due to the who bring into play cell phones at the same time as driving, drink and drive, etc. Due to this, most of the companies of automobiles tries to make available best Advanced Driver Assistance System (ADAS) to the customer to avoid the accidents. The lane detection approach is one of the method provided by automobile companies in ADAS, in which the vehicle must follows the lane. Therefore, there is less chance to get an accident. The information obtained from the lane is used to alert the driver. Therefore most of the researchers are attracted towards this field. But, due to the varying road circumstances, it is very difficult to detect the lane. The computer apparition and machine learning approaches are presents in most of the articles. In this article, we presents the deep learning scheme for identification of lane. There are two phases are presents in this work. In a first phase the image transformation is done and in second phase lane detection is occurred. At first, the proposed model gets the numerous lane pictures and changes the picture into its relating Bird's eye view picture by using Inverse perspective mapping transformation. The Deep Convolutional Neural Network (DCNN) classifier to identify the lane from the bird’s eye view image. The Earth Worm- Crow Search Algorithm (EW-CSA) is designed to help DCNN with the optimal weights. The DCNN classifier gets trained with the view picture from the bird’s eye image and the optimal weights are selected through newly developed EW-CSA algorithm. All these algorithms are performed in MATLAB. The simulation results shows that the exact detection of lane of road. Also, the accuracy, sensitivity, and specificity are calculated and its values are 0.99512, 0.9925, and 0.995 respectively.

SLNet: Stereo face liveness detection via dynamic disparity-maps and convolutional neural network

Current state-of-the-art dual camera-based face liveness detection methods utilize either hand-crafted features, such as disparity, or deep texture features to classify a live face and face Presentation Attack (PA). However, these approaches limit the effectiveness of classifiers, particularly deep Convolutional Neural Networks (CNN) to unknown face PA in adverse scenarios. In contrast to these approaches, in this paper, we show that supervising a deep CNN classifier by learning disparity features using the existing CNN layers improves the performance and robustness of CNN to unknown types of face PA. For this purpose, we propose to supervise a CNN classifier by introducing a disparity layer within CNN to learn the dynamic disparity-maps. Subsequently, the rest of the convolutional layers, following the disparity layer, in the CNN are supervised using the learned dynamic disparity-maps for face liveness detection. We further propose a new video-based stereo face anti-spoofing database with various face PA and different imaging qualities. Experiments on the proposed stereo face anti-spoofing database are performed using various test case scenarios. The experimental results indicate that our proposed system shows promising performance and has good generalization ability.

Handwritten Kannada Numerals Recognition by Using Zone Features and CNN Classifier

The proposed system presents a pre-processing, segmentation, features extraction approach and Deep Convolutional Neural Network (DCNN) classifier for recognition of handwritten Kannada numerals. Pre-processing have different steps like median filter, gray scale to binary, normalization, thinning, skew correction and slant removal. Segmentation process contains different methods like vertical projection profile for word and novel character segmentation. Collections of best discriminable features are very important part in achieving high rate of identification in automatic numeral detection systems. Kannada is the major south Indian character verbal by about 50 million people. This article presents a well-organized and novel technique for recognition of handwritten Kannada numerals using zone and distance matrix. An appropriate feature extractor and a superior classifier play most important task in achieving high detection rate for a recognition scheme. This article determines a variety of feature extraction approaches and classification techniques which are designed to recognize handwritten numerals of Kannada script. The DCNN classifier approach is used to classify the testing samples of each Kannada handwritten numerals. The experimental result gives the acceptable performance rate.

Progressive Transfer Learning and Adversarial Domain Adaptation for Cross-Domain Skin Disease Classification.

Deep learning has been used to analyze and diagnose various skin diseases through medical imaging. However, recent researches show that a well-trained deep learning model may not generalize well to data from different cohorts due to domain shift. Simple data fusion techniques such as combining disease samples from different data sources are not effective to solve this problem. In this paper, we present two methods for a novel task of cross-domain skin disease recognition. Starting from a fully supervised deep convolutional neural network classifier pre-trained on ImageNet, we explore a two-step progressive transfer learning technique by fine-tuning the network on two skin disease datasets. We then propose to adopt adversarial learning as a domain adaptation technique to perform invariant attribute translation from source to target domain in order to improve the recognition performance. In order to evaluate these two methods, we analyze generalization capability of the trained model on melanoma detection, cancer detection, and cross-modality learning tasks on two skin image datasets collected from different clinical settings and cohorts with different disease distributions. The experiments prove the effectiveness of our method in solving the domain shift problem.

A Hybrid Convolutional Neural Network and Deep Belief Network for Brain Tumor Detection in MR Images

Early tumor detection in brain plays vital role in early tumor detection and radiotherapy. MR images are used as the input image for brain tumor finding and classify the type of brain tumor. For early detection or prediction of the brain tumor, an improved feature extraction technique along with Deep Neural Network (DNN) has been recommended. First, MR image is pre-processed, segmented and classified utilizing image processing techniques. Support Vector Machine (SVM) based brain tumor classifications are achieved previously with less precision rate. By integrating DCNN(Deep Convolutional Neural Network) classifier and DBN(Deep Belief Network), an improvement in precision rate can be achieved. This paper mainly focuses on six features viz., entropy, mean, correlation, contrast, energy and homogeneity. The proposed method is used to identify the place, locality and dimension (size) of the tumor in the cerebrum through MR copy using MATLAB software. The performance metrics recall, precision, sensitivity, accuracy and specificity are achieved.

Multi-modal classifier fusion with feature cooperation for glaucoma diagnosis

ABSTRACTGlaucoma is a major public health problem that can lead to an optic nerve lesion, requiring systematic screening in the population over 45 years of age. The diagnosis and classification of this disease have had a marked and excellent development in recent years, particularly in the machine learning domain. Multimodal data have been shown to be a significant aid to the machine learning domain, especially by its contribution to improving data driven decision-making.Solving classification problems by combinations of classifiers has made it possible to increase the robustness as well as the classification reliability by using the complementarity that may exist between the classifiers. Complementarity is considered a key property of multimodality. A Convolutional Neural Network (CNN) works very well in pattern recognition and has been shown to exhibit superior performance, especially for image classification which can learn by themselves useful features from raw data. This article proposes a multimodal classification approach based on deep Convolutional Neural Network and Support Vector Machine (SVM) classifiers using multimodal data and multimodal feature for glaucoma diagnosis from retinal fundus images from RIM-ONE dataset. We make use of handcrafted feature descriptors such as the Gray Level Co-Occurrence Matrix, Central Moments and Hu Moments to co-operate with features automatically generated by the CNN in order to properly detect the optic nerve and consequently obtain a better classification rate, allowing a more reliable diagnosis of glaucoma.The experimental results confirm that the combination of classifiers using a new hybrid fusion approach and the BWWV technique is better than learning classifiers separately. The proposed method provides a computerized diagnosis system for glaucoma disease with impressive results comparing them to the main related studies that allow us to continue in this research path.

Lung cancer diagnosis with quantitative DIC microscopy and a deep convolutional neural network.

We present a study on lung squamous cell carcinoma diagnosis using quantitative TI-DIC microscopy and a deep convolutional neural network (DCNN). The 2-D phase map of unstained tissue sections is first retrieved from through-focus differential interference contrast (DIC) images based on the transport of intensity equation (TIE). The spatially resolved optical properties are then computed from the 2-D phase map via the scattering-phase theorem. The scattering coefficient ( ) and the reduced scattering coefficient ( ) are found to increase whereas the anisotropy factor (g) is found to decrease with cancer. A DCNN classifier is developed afterwards to classify the tissue using either the DIC images or 2-D optical property maps of , and g. The DCNN classifier with the optical property maps exhibits high accuracy, significantly outperforming the same DCNN classifier on the DIC images. The label-free quantitative phase microscopy together with deep learning may emerge as a promising approach for in situ rapid cancer diagnosis.

A Novel CNN-Based Framework for Classification of Signal Quality and Sleep Position from a Capacitive ECG Measurement.

The further exploration of the capacitive ECG (cECG) is hindered by frequent fluctuations in signal quality from body movement and changes in sleep position. The processing framework must be fundamentally adapted to make full use of this signal. Therefore, we propose a new signal-processing framework that determines the signal quality for short signal segments (2 and 4 seconds) using a multi-class classification model (qua_model) based on a convolutional neural network (CNN). We built another independent deep CNN classifier (pos_model) to classify the sleep position. In the validation, 12 subjects were recruited for a 30-minute experiment, which required the subjects to lie on a bed in different sleeping positions. The short segments, classified as clear (C1 class) by the qua_model, were used to determine sleep positions with the pos_model. In 10-fold cross-validation, the qua_model for signals of 4-second length could recognize the signal of the C1 class at a 0.99 precision and a 0.99 recall; the pos_model could recognize the supine sleep position, the left, and right lateral sleep positions at a 0.99 averaged precision and a 0.99 averaged recall. Given the amount of data accumulated per night and the instability in the signal quality, this fully automatic processing framework is indispensable for a personal healthcare system. Therefore, this study could serve as an important step for cECG technique trying to explore the cECG for unconstrained heart monitoring.

Glioma Tumor Grade Identification Using Artificial Intelligent Techniques

Computer aided diagnosis using artificial intelligent techniques made tremendous improvement in medical applications especially for easy detection of tumor area, tumor type and grades. This paper presents automatic glioma tumor grade identification from magnetic resonant images using Wndchrm tool based classifier (Weighted Neighbour Distance using Compound Heirarchy of Algorithms Representing Morphology) and VGG-19 deep convolutional neural network (DNN). For experimentation, DICOM images are collected from reputed government hospital and the proposed intelligent system categorized the tumor into four grades such as low grade glioma, oligodendroglioma, anaplastic glioma and glioblastoma multiform. After preprocessing, features are extracted, optimized and then classified using Windchrm tool where the most significant features are selected on the basis of Fisher score. In the case of DNN classifier, data augmentation is also performed before applying the images into the deep learning network. The performance of the classifiers are analysed with various measures such as accuracy, precision, sensitivity, specificity and F1-score. The results showed reasonably good performance with a maximum classification accuracy of 92.86% for the Wndchrm classifier and 98.25% for VGG-19 DNN classifier. The results are also compared with similar recent works and the proposed system is found to have better performance.

RETRACTED ARTICLE: Refining Parkinson’s neurological disorder identification through deep transfer learning

Parkinson’s disease (PD), a multi-system neurodegenerative disorder which affects the brain slowly, is characterized by symptoms such as muscle stiffness, tremor in the limbs and impaired balance, all of which tend to worsen with the passage of time. Available treatments target its symptoms, aiming to improve the quality of life. However, automatic diagnosis at early stages is still a challenging medicine-related task to date, since a patient may have an identical behavior to that of a healthy individual at the very early stage of the disease. Parkinson’s disease detection through handwriting data is a significant classification problem for identification of PD at the infancy stage. In this paper, a PD identification is realized with help of handwriting images that help as one of the earliest indicators for PD. For this purpose, we proposed a deep convolutional neural network classifier with transfer learning and data augmentation techniques to improve the identification. Two approaches like freeze and fine-tuning of transfer learning are investigated using ImageNet and MNIST dataset as source task independently. A trained network achieved 98.28% accuracy using fine-tuning-based approach using ImageNet and PaHaW dataset. Experimental results on benchmark dataset reveal that the proposed approach provides better detection of Parkinson’s disease as compared to state-of-the-art work.