Evaluating Convolutional Neural Networks Reliability depending on their Data Representation

Objects and processes classification is a common experimental problem. Its solution, first of all, is needed in automatic diagnosis systems, for example, to determine the equipment operation state through the diagnostic signal or to identify abnormalities with medical images. With the development of convolutional neural networks, new prospects for solving such problems have risen up. However, the classification accuracy that can be achieved on these networks is not sufficient enough for all diagnosis issues. It is subject to, for example, timely diagnosis of the onset of transient phenomena. At the same time, another type of neural network, Kohonen self-organizing maps, has a conceptual property for training on unclassified set of classes, that is, giving the opportunity to solve such an issue. Therefore, the accuracy enhancement of the classification on the basis of Kohonen networks implementation in the architecture of convolutional networks is a relevant objective and has practical significance. The article analyzes the means of improving the accuracy of convolutional neural networks and arising problems solution. The ways of increasing the proportion of correct clustering on Kohonen networks due to the ‘growth’ of its grid shape in determining new classes in the learning process are also given. It is this property that makes it possible to recognize transient phenomena. It is determined that the existing solutions of the combination of Kohonen networks and convolutional networks are aimed at improving the efficiency of only self-organizing maps, with the purpose to improve the accuracy of classification by convolutional networks, it became necessary to develop a new architecture. The paper provides a description of this issue. Since the initial information of the Kohonen networks is the weighting matrix values of the grid shape neurons, it was necessary to associate it with the representation of the images in order to process the diagnostic images. The paper proposes the concept of a built-in associative array block based on Kohonen networks. According to the proposed method, a software implementation of a hybrid neural network is developed. The formulation and results of computational experiments are presented. The efficiency of the proposed method is experimentally proved. Keywords: convolutional neural networks, CNN, Kohonen self-organizing maps, SOM, signal classification, image classification.

Cancer is now the leading causes of death. During treatment, if an abnormal mass is detected by routine mammography after histological examination, a biopsy will be performed for further diagnosis. However, the time required to review and evaluate the biopsy slide is very long. An effective algorithm that can identify cancer tissues and reduce the rate of misdiagnosis is urgently needed. The current research methods mainly include: Capsule Network, and Deep Convolutional Neural Networks (CNNs) used in image classification, etc. This paper mainly uses Capsule Network, CNN, CNN trained on Histogram of Oriented Gradient (HOG) images, and Grey Level Co-occurrence Matrix (GLCM) to classify benign and malignant images of the tumor micro-biopsy images. The first step is to download the breast tumor micro-biopsy images from Breakhis. The second step is to preprocess the images, mainly image enhancement, augmentation, and HOG feature extraction. In the third step, the processed image data is used to train CNN and Capsule Network. The last step is to visualize the accuracy of model training and use confusion matrix to calculate the accuracy of image classification, recall rate, and so on. The results show that the accuracy of CNN with the help of transfer learning can reach 98.32%, the accuracy of CNN trained with HOG image is only 82.99%, the accuracy of CNN trained with GLCM is 92.79%, and the accuracy of capsule network can reach 92.39%.

Fine-tuning Convolutional Neural Networks for fine art classification

Convolutional neural networks (CNNs) are being used in this paper to investigate waste classification. The study compared the categorization of three waste categories (plastic, paper, and metal) using a CNN and a Support Vector Machine (SVM). The results showed that while the SVM had an accuracy of 94.8%, the CNN's accuracy was lower at 83%. Another study divided waste into six categories and showed that the classification accuracy was 22% with a CNN, while the SVM had a test accuracy of 63%. The authors noted challenges in optimizing the hyperparameters, which prevented the CNN from being trained to its full potential. However, they concluded that the CNN should attain higher accuracy levels than the SVM once these challenges are overcome. The authors also used two popular CNN architectures, VGG16 and FastNet-34, in their proposed model for waste classification. The dataset used for the model was provided by Kaggle and consisted of 22564 waste photos divided into recyclable (R) and organic (O) categories. The results of the proposed model showed improved accuracy in waste classification compared to previous studies. The study highlights the potential for the use of CNNs in the field of waste management but highlights the importance of overcoming challenges in hyperparameter optimization for higher computer vision accuracy.

PurposeThe aim of this study was to analyse and review deep learning convolutional neural networks for detecting and diagnosing early-stage dental caries on periapical radiographs.Materials and MethodsIn order to conduct this review, the Preferred Reporting Items for Systematic Reviews and Meta-Analyses (PRISMA) guidelines were followed. Studies published from 2015 to 2021 under the keywords (deep convolutional neural network) AND (caries), (deep learning caries) AND (convolutional neural network) AND (caries) were systematically reviewed.ResultsWhen dental caries is improperly diagnosed, the lesion may eventually invade the enamel, dentin, and pulp tissue, leading to loss of tooth function. Rapid and precise detection and diagnosis are vital for implementing appropriate prevention and treatment of dental caries. Radiography and intraoral images are considered to play a vital role in detecting dental caries; nevertheless, studies have shown that 20% of suspicious areas are mistakenly diagnosed as dental caries using this technique; hence, diagnosis via radiography alone without an objective assessment is inaccurate. Identifying caries with a deep convolutional neural network-based detector enables the operator to distinguish changes in the location and morphological features of dental caries lesions. Deep learning algorithms have broader and more profound layers and are continually being developed, remarkably enhancing their precision in detecting and segmenting objects.ConclusionClinical applications of deep learning convolutional neural networks in the dental field have shown significant accuracy in detecting and diagnosing dental caries, and these models hold promise in supporting dental practitioners to improve patient outcomes.

This article presents the artificial intelligence for object recognition by touching method with the tactile sensor, which has similar structural characteristics as the human hand. The object recognition uses a tactile glove sensor with small tactile sensors to spread throughout the glove, and there are touching points similar to human hands. Using the glove sensor to grasp the object that will obtain an image, there is tactile image. These images were used to compare recognition performance with two techniques: the Bag of Word (BoW) and the Convolutional Neural Network (CNN). In the experiment, the tactile glove performs 20 different objects. The BoW technique, the Scale Invariant Feature Transform (SIFT), has been used to feature extraction and then classified with K Nearest Neighbors (KNN) to evaluate the performance. The results illustrate that the accuracy of BoW and CNN is 70.80% and 97.20%, respectively. When comparing both techniques, it was found that the CNN had higher accuracy of about 26.40%. The results clearly show that CNN was more performance than BoW. Therefore, it is suitable to analyze tactile glove recognition, which can be applied to the recognition system of humanoid robots.

The perceived texture directionality is an important, not fully explored image characteristic. In many applications texture directionality detection is of fundamental importance. Several approaches have been proposed, such as the fast Fourier-based method. We recently proposed a method based on the interpolated grey-level co-occurrence matrix (iGLCM), robust to image blur and noise but slower than the Fourier-based method. Here we test the applicability of convolutional neural networks (CNNs) to texture directionality detection. To obtain the large amount of training data required, we built a training dataset consisting of synthetic textures with known directionality and varying perturbation levels. Subsequently, we defined and tested shallow and deep CNN architectures. We present the test results focusing on the CNN architectures and their robustness with respect to image perturbations. We identify the best performing CNN architecture, and compare it with the iGLCM, the Fourier and the local gradient orientation methods. We find that the accuracy of CNN is lower, yet comparable to the iGLCM, and it outperforms the other two methods. As expected, the CNN method shows the highest computing speed. Finally, we demonstrate the best performing CNN on real-life images. Visual analysis suggests that the learned patterns generalize to real-life image data. Hence, CNNs represent a promising approach for texture directionality detection, warranting further investigation.

A genetic algorithm (GA), is used to optimize the many parameters of a convolutional neural network (CNN) that control the structure of the network. CNNs are used in image classification problems where it is necessary to generate feature descriptors to discern between image classes. Because of the deep representation of image data that CNNs are capable of generating, they are increasingly popular in research and industry applications. With the increasing number of use cases for CNNs, more and more time is being spent to come up with optimal CNN structures for different applications. Where one CNN might succeed at classification, another can fail. As a result, it is desirable to more easily find an optimal CNN structure to increase classification accuracy. In the proposed method, a GA is used to evolve the parameters that influence the structure of a CNN. The GA compares CNNs by training them on images of concrete containing cracks. The best CNN after several generations of the GA is then compared to the state-of-the-art CNN for crack detection. This work shows that it is possible to generalize the process of optimizing a CNN for image classification through the use of a GA.

Convolutional neural networks (CNNs) have been successfully applied to many tasks such as digit and object recognition. In this paper we study the size of the receptive field of deep convolutional neural networks, in particular, we check the idea of a redundant receptive field. We run a set of experiments on two common CNN models — VGG16 and ResNet18 — in order to explore the influence of receptive field size on CNN's training time, accuracy, and performance. We run experiments using the MakiFlow framework on the CALTECH256 dataset. The experiments' results show that the optimization of neural networks (NNs) by reducing the size of the receptive field allows to reduce the NN’s training time by 5-7% while maintaining the accuracy of the network.

BackgroundMachine learning studies require a large number of images often obtained on different PET scanners. When merging these images, the use of harmonized images following EARL-standards is essential. However, when including retrospective images, EARL accreditation might not have been in place. The aim of this study was to develop a convolutional neural network (CNN) that can identify retrospectively if an image is EARL compliant and if it is meeting older or newer EARL-standards.Materials and methods96 PET images acquired on three PET/CT systems were included in the study. All images were reconstructed with the locally clinically preferred, EARL1, and EARL2 compliant reconstruction protocols. After image pre-processing, one CNN was trained to separate clinical and EARL compliant reconstructions. A second CNN was optimized to identify EARL1 and EARL2 compliant images. The accuracy of both CNNs was assessed using fivefold cross-validation. The CNNs were validated on 24 images acquired on a PET scanner not included in the training data. To assess the impact of image noise on the CNN decision, the 24 images were reconstructed with different scan durations.ResultsIn the cross-validation, the first CNN classified all images correctly. When identifying EARL1 and EARL2 compliant images, the second CNN identified 100% EARL1 compliant and 85% EARL2 compliant images correctly. The accuracy in the independent dataset was comparable to the cross-validation accuracy. The scan duration had almost no impact on the results.ConclusionThe two CNNs trained in this study can be used to retrospectively include images in a multi-center setting by, e.g., adding additional smoothing. This method is especially important for machine learning studies where the harmonization of images from different PET systems is essential.

Convolutional Neural Networks (CNNs) have revolutionized the world of image classification over the last few years, pushing the computer vision close beyond human accuracy. The required computational effort of CNNs today requires power-hungry parallel processors and GP-GPUs. Recent efforts in designing CNN Application-Specific Integrated Circuits (ASICs) and accelerators for System-On-Chip (SoC) integration have achieved very promising results. Unfortunately, even these highly optimized engines are still above the power envelope imposed by mobile and deeply embedded applications and face hard limitations caused by CNN weight I/O and storage. On the algorithmic side, highly competitive classification accuracy canbe achieved by properly training CNNs with binary weights. This novel algorithm approach brings major optimization opportunities in the arithmetic core by removing the need for the expensive multiplications as well as in the weight storage and I/O costs. In this work, we present a HW accelerator optimized for BinaryConnect CNNs that achieves 1510 GOp/s on a corearea of only 1.33 MGE and with a power dissipation of 153 mW in UMC 65 nm technology at 1.2 V. Our accelerator outperforms state-of-the-art performance in terms of ASIC energy efficiency as well as area efficiency with 61.2 TOp/s/W and 1135 GOp/s/MGE, respectively.

PurposeTo compare the diagnostic accuracy of convolutional neural networks (CNN) with radiologists as the reference standard in the diagnosis of intracranial hemorrhages (ICH) with non contrast computed tomography of the cerebrum (NCTC). MethodsPubMed, Embase, Scopus, and Web of Science were searched for the period from 1 January 2012 to 20 July 2020; eligible studies included patients with and without ICH as the target condition undergoing NCTC, studies had deep learning algorithms based on CNNs and radiologists reports as the minimum reference standard. Pooled sensitivities, specificities and a summary receiver operating characteristics curve (SROC) were employed for meta-analysis. Results5,119 records were identified through database searching. Title-screening left 47 studies for full-text assessment and 6 studies for meta-analysis. Comparing the CNN performance to reference standards in the retrospective studies found a pooled sensitivity of 96.00% (95% CI: 93.00% to 97.00%), pooled specificity of 97.00% (95% CI: 90.00% to 99.00%) and SROC of 98.00% (95% CI: 97.00% to 99.00%), and combining retrospective and studies with external datasets found a pooled sensitivity of 95.00% (95% CI: 91.00% to 97.00%), pooled specificity of 96.00% (95% CI: 91.00% to 98.00%) and a pooled SROC of 98.00% (95% CI: 97.00% to 99.00%). ConclusionThis review found the diagnostic performance of CNNs to be equivalent to that of radiologists for retrospective studies. Out-of-sample external validation studies pooled with retrospective studies found CNN performance to be slightly worse. There is a critical need for studies with a robust reference standard and external data-set validation.

The use of machine learning for the image classification process is growing all the time. Many methods can be used to classify an image with good accuracy. Convolutional Neural Network (CNN) and Support Vector Machine (SVM) are popular methods for this case. The two approaches have differences in the data training process to achieve classification objectives. Although there are some differences between these approaches, there are some advantages to both of them. This research explores the comparison of the two CNN and SVM methods by comparing the training process carried out and the accuracy results of the classification. The process stages are divided into pre-processing, training, and testing. The objects used are ten waste plastic bottles with different brands of medium size with a total data of 1100 images. Based on the observations, both methods have advantages and disadvantages in the data training and classification process. However, from the results, CNN's accuracy is better than SVM. The accuracy of both networks is 99% for CNN and 74% for SVM, respectively. So, from the results of experiments that have been carried out in the study, it was found that CNN was still better than SVM. Doi: 10.28991/ESJ-2022-06-05-011 Full Text: PDF

The report presents an algorithm for constructing a convolutional hierarchical neural network classifier, which is a modification of the algorithm for constructing hierarchical neural network classifiers suggested before. The original algorithm was designed to exploit intrinsic class hierarchy to build a class tree with a neural network in each node classifying groups of initial classes (in a non-terminal node) or a subset of original classes (in a terminal node). The convolutional modification utilizes convolutional neural networks instead of regular fully connected networks in order to apply the model to image classification tasks. Use of class hierarchy for image classification should reduce the number of adjusted neural network parameters compared to deep convolutional neural networks, and therefore it should reduce training and inference time. In this context the algorithm may be compared with some pruning techniques. The convolutional hierarchical neural network classifier inherits some hyperparameters of a conventional hierarchical neural network classifier, like the activation threshold and the threshold by the share of voting patterns. The goal of this study was to explore different strategies of choosing these hyperparameters. To test these strategies, we used the CIFAR-10 dataset. Also, for demonstration purposes we apply the convolutional hierarchical neural network classifier to the CIFAR-100 dataset.

Hyperspectral Image (HSI) classification is one of the fundamental tasks in the field of remote sensing data analysis. CNN (Convolutional Neural Network) has been proven to be an effective deep learning model, which can extract high-level features directly from the raw data and thereby utilize rich information contained in HSI data. However, labor cost to label enough HIS data for training model is usually expensive, so that it is a strong demand of utilizing limited training data to get a satisfied classification accuracy. In this paper, we put forward a deep cube CNN model – DCNR, which is composed of a cube neighbor HSI pixels strategy, a deep CNN and a random forest classifier. In DCNR model, cubic samples, containing spectral-spatial information, are generated by putting each target pixel and its neighbors together. Then features with high representative ability, extracted by applying a specially designed cube CNN model on each cubic sample, are fed into the random forest classifier for the classification of the target pixel. Results show that DCNR model can achieve classification accuracy of 96.78%, 96.08% and 94.85% on KSC, IP and SA datasets respectively with 20% samples as training set, and 85.03%, 83.45 and 62.17% on KSC, IP and SA datasets respectively with only 1% samples as training set, significantly outperforming random forest and cube CNN models.