Network For Image Recognition Research Articles

A Multi-task Neural Network for Image Recognition in Magnetically Controlled Capsule Endoscopy.

Physicians are required to spend a significant amount of reading time of magnetically controlled capsule endoscopy. However, current deep learning models are limited to completing a single recognition task and cannot replicate the diagnostic process of a physician. This study aims to construct a multi-task model that can simultaneously recognize gastric anatomical sites and gastric lesions. A multi-task recognition model named Mul-Recog-Model was established. The capsule endoscopy image data from 886 patients were selected to construct a training set and a test set for training and testing the model. Based on the same test set, the model in this study was compared with the current single-task recognition model with good performance. The sensitivity and specificity of the model for recognizing gastric anatomical sites were 99.8% (95% confidence intervals: 99.7-99.8) and 98.5% (95% confidence intervals: 98.3-98.7), and for gastric lesions were 98.8% (95% confidence intervals: 98.3-99.2) and 99.4% (95% confidence intervals: 99.1-99.7). Moreover, the positive predictive value, negative predictive value, and accuracy of the model were more than 95% in recognizing gastric anatomical sites and gastric lesions. Compared with the current single-task recognition model, our model showed comparable sensitivity, specificity, positive predictive value, negative predictive value, and accuracy (p < 0.01, except for the negative predictive value of ResNet, p > 0.05). The Areas Under Curve of our model were 0.985 and 0.989 in recognizing gastric anatomical sites and gastric lesions. Furthermore, the model had 49.1M parameters and 38.1G Float calculations. The model took 15.5ms to recognize an image, which was less than the superposition of multiple single models (p < 0.01). The Mul-Recog-Model exhibited high sensitivity, specificity, PPV, NPV, and accuracy. The model demonstrated excellent performance in terms of parameters quantity, Float computation, and computing time. The utilization of the model for recognizing gastric images can improve the efficiency of physicians' reports and meet complex diagnostic requirements.

A lightweight visual mamba network for image recognition under resource-limited environments

The Vision Transformers (ViTs) models show great potential in recognition due to their excellent self-attention mechanisms. However, they often need more computational complexity, making achieving high performance in resource-constrained environments challenging. This paper proposes a lightweight visual model (called LightViM) to tackle these challenges, specifically devised for recognition challenges under resource-constrained environments. Considering the imperative of reducing the model’s computational complexity and resource consumption , we propose lightweight local–global feature fusion modules based on Mamba (LGF-Mamba) aimed at integrating spatially detailed local information with globally contextualized features while maintaining lower linear time complexity. In LGF-Mamba, Mamba sub-blocks are first designed to extract global information; then, for the rapid extraction of local features, LocalE module is designed and integrated into LGF-Mamba, exploring more comprehensive feature representation. This mechanism efficiently directs the network to integrate features from different spatial scales, thereby improving the recognition performance in resource-constrained environments. Experimental results indicate that, in comparison to other leading-edge lightweight methods, the proposed procedure simultaneously achieves superior recognition performance and minimal resource consumption.

Bio-inspired deep neural local acuity and focus learning for visual image recognition

In the field of computer vision and image recognition, enabling the computer to discern target features while filtering out irrelevant ones poses a challenge. Drawing insights from studies in biological vision, we find that there is a local visual acuity mechanism and a visual focus mechanism in the initial conversion and processing of visual information, ensuring that the visual system can give ear to salient features of the target in the early visual observation phase. Inspired by this, we build a novel image recognition network to focus on the target features while ignoring other irrelevant features in the image, and further focus on the focus features based on the target features. Meanwhile, in order to comply with the output characteristics when similar features exist in different categories, we design a softer image label operation for similar features in different categories, which solves the correlation of labels between categories. Relevant experimental findings underscore the efficacy of our proposed method, revealing discernible advantages in comparison to alternative approaches. Visualization results further attest to the method’s capability to selectively focus on pertinent target features within the image, sidelining extraneous information.

Application of Convolutional Neural Network in Image Recognition

Image recognition is of great significance for both human social development and daily life. This paper mainly discusses the application research of convolutional neural network (CNN) in the field of image recognition. Firstly, the basic structure and working principle of CNN, including the core components of convolution layer, pooling layer and full connection layer, are introduced, and then the structure and characteristics of the middle layer are explained. Then, this paper compares the advantages and disadvantages of different CNN models such as LeNet-5, AlexNet, GoogLeNet, ResNet, etc., and analyzes their application in face recognition, medical image recognition, traffic recognition, character recognition and other fields. Finally, this paper summarizes the application of CNN in image recognition, and points out the problems faced in its practical application, such as model complexity, computational resource requirements, etc., aiming to provide reference for further research of image recognition technology.

Integration of Ag-based threshold switching devices in silicon microchips

Threshold-type resistive switching (RS) is an essential electronic behavior in many types of integrated circuits and can be exploited in multiple applications, such as leaky integrate-and-fire neurons for artificial neural networks (ANNs). Many research articles have shown that metal/insulator/metal (MIM) devices using Ag electrodes exhibit stable threshold-type RS, but all of them presented large devices (>1 µm2) fabricated on unfunctional SiO2/Si substrates. In this article, for the first time we integrate Ag-based threshold-type RS devices at the back-end-of-line interconnections of silicon microchips. The insulator used is multilayer hexagonal boron nitride (h-BN), and the size of the devices is ∼0.05 µm2. The devices can switch between a high resistive state (HRS) and a low resistive state (LRS) without the need of any forming process, and we observe a high endurance over 1 million cycles over multiple devices. By performing circuit simulation using SPICE software, we confirm that this electrical behavior is suitable for being used as leaky integrate-and-fire electronic neuron in spiking neural networks for image recognition, and the h-BN artificial neurons operate correctly for 94 % of the images presented. Our study represents a significant advancement towards the integration of Ag-based threshold-type RS devices in silicon microchips.

Lightweight Food Recognition via Aggregation Block and Feature Encoding

Food image recognition has recently been given considerable attention in the multimedia field in light of its possible implications on health. The characteristics of the dispersed distribution of ingredients in food images put forward higher requirements on the long-range information extraction ability of neural networks, leading to more complex and deeper models. Nevertheless, the lightweight version of food image recognition is essential for improved implementation on end devices and sustained server-side expansion. To address this issue, we present Aggregation Feature Net(AFNet), a lightweight network that is capable of effectively capturing both global and local features from food images. In AFNet, we develop a novel convolution based on a residual model by encoding global features through row-wise and column-wise information integration. Merging aggregation block with classic local convolution yields a framework that works as the backbone of the network. Based on the efficient use of parameters by the aggregation block, we constructed a lightweight food image recognition network with fewer layers and a smaller scale, assisted by a new type of activation function. Experimental results on four popular food recognition datasets demonstrate that our approach achieves state-of-the-art performance with higher accuracy and fewer FLOPs and parameters. For example, in comparison to the current state-of-the-art model of MobileViTv2, AFNet achieved 88.4% accuracy of the top-1 level on the ETHZ Food-101 dataset, with similar parameters and FLOPs but 1.4% more accuracy. The source code will be provided in supplementary materials.

Constructing a cataplexy face prediction model for narcolepsy type 1 based on ResNet-18

Objective: To establish a prediction model for the identifying of cataplexy facial features based on clinical shooting videos by using a deep learning image recognition network ResNet-18. Methods: A cross-sectional study. Twenty-five narcolepsy type 1 patients who were first diagnosed and never received treatment and 25 healthy controls recruited by advertisement in the Second Affiliated Hospital of Nanchang University from 2020 to 2023.After image preprocessing, a total of 1 180 images were obtained, including 583 cataplexy faces and 597 normal faces.90% were selected as the training set and validation set, and then expanded the data by 5 times.80% of the expanded data set was extracted as the training set and 20% as the validation set, that is, the number of the training set was (583+597)×0.9×0.8×5=4 248, the number of the validation set was (583+597)×0.9×0.2×5=1 062. The data sets for training and validation were used train parameters to establish the model and were trained through the five-fold cross-validation method, to establish the ResNet-18 cataplexy face recognition model via transfer learning.10% (118 images) of the original non-amplified images were extracted as the test set. The test set data did not participate in data enhancement and model training, and was only used to evaluate the final performance of the model. Finally, ResNet-18 was compared with VGG-16, ResNet-34 and Inception V3 deep learning models, and the receiver operating characteristic curve was used to evaluate the value of ResNet-18 image recognition network in cataplexy face recognition. Results: Among 25 patients with narcolepsy type 1, 15 were males and 10 were females, aged [M (Q1, Q3)] of 14.0(11.0, 20.5) years.Among 25 healthy controls, 14 were males and 11 were females, with a median age of 16.0(14.4, 23.0) years.The overall accuracy of ResNet-18 image recognition network in the test set was 90.9%, the sensitivity was 96.4% and the specificity was 85.2%. The area under the ROC curve was 0.99(95%CI:0.96-1.00). The ResNet-18 model parameter amount was 11.69 M, the floating point operation amount was 1 824.03 M, and the single image recognition time was 5.9 ms. Conclusions: The cataplexy face prediction model built based on the deep learning image recognition network ResNet-18 has a high accuracy in identifying cataplexy faces.

Water transport in PEMFC with metal foam flow fields: Visualization based on AI image recognition

Water transport is an essential process in proton exchange membrane fuel cells (PEMFC). For the novel metal foam flow fields, water transport is difficult to be visualized and still remains inadequately understood. In this work, for the first time, the two-phase flows in the metal foam flow fields of a transparent PEMFC was visualized by using an artificial intelligence (AI) image recognition method. In order to distinguish water from the metal foam ligaments, a scoring system was developed, which effectively solved the problem of inconsistent reflectance of liquid water between the ligament and the gas diffusion layer. According to the state of liquid water, the scoring system categorized the local images into three labels of waterless state, wet state, and slug state for AI image recognition network training. The trained AI image recognition model was then used to visualize the liquid water state in the metal foam flow fields under different operating conditions of relative humidity (RH), temperature, and inlet flow rate. Results show that, since the porous structure of the metal foam provides flexible transport paths, the PEMFC can operate effectively in a wide range of wet states. Even a 25% difference in the wet-state percentage between different operating conditions yields an almost consistent voltage. The increased slugs in the flow fields are the primary reason for the performance degradation of the PEMFC. At a high current density of 2 A/cm2, the voltage is as low as 0.03 V when the slug state occupies 35% of the flow fields (100% RH, 60 °C, 1.0 SLPM cathode inlet flow rate), and as high as 0.56 V when the slug state is absent in the flow fields (100% RH, 60 °C, 2.0 SLPM cathode inlet flow rate). The present work provides a novel approach for the studies of water transport in PEMFC.

Performance of convolutional neural networks on CIFAR and Dog Breed classification dataset

This research aims to study the deep learning applications in image recognition and classification tasks. The advantages and limitations of deep learning are explored by analysing existing deep learning algorithms and their applications on tasks such as using images to do classification, image segmentation, and target detection. In the experimental part, this paper evaluates the performance of deep learning in image recognition by using classical neural network models, including training and testing models on a large size of image datasets. The results show that neural network has good classification and detection capabilities in image recognition tasks and also achieves good results for image segmentation tasks. However, the model training process of shallow neural network models is time-consuming and performs poorly for small-scale datasets. Based on these situations, this paper proposes some optimization strategies to achieve high performance and efficiency of neural networks in image recognition and analyze the performance differences between different strategies .

Enhanced Memristive Performance via a Vertically Heterointerface in Nanocomposite Thin Films for Artificial Synapses.

Memristors can be used to mimic synaptic behavior in artificial neural networks, which makes them a key component in neuromorphic computing and holds promise for advancing the field. In this study, a memory artificial synaptic device based on ZnO-BaTiO3 (ZnO-BTO) vertically aligned nanocomposite thin films was prepared. The vertical interface between the two phases can be used as a conduit for oxygen vacancy (OV) accumulation and a channel for OV movement, which greatly optimizes the resistive switching performance of the device and has the potential for multistage storage. By applying different pulse sequences to the device, the conductance of the device is adjusted from multiple angles, and a variety of synaptic functions are simulated, such as paired-pulse facilitation, spike-timing-dependent plasticity, short-term plasticity to long-term plasticity (STP-LTP), and long-term potentiation/depression (LTP/LTD). Finally, we construct a neural network for image recognition, and the recognition accuracy can reach 91%. Our study demonstrates the feasibility of using composite thin-film vertical interface to regulate the resistive performance of memristors and its great potential in artificial synaptic simulation and neuromorphic computing.

Guest Editorial: Foreword to the Special Issue on Advanced Research and Applications of Deep Learning and Neural Network in Image Recognition

Over the last two decades, the realm of image recognition has undergone a remarkable transformation, characterized by an astonishing pace of advancement [...]

Point of Sale System Using Convolutional Neural Network for Image Recognition in Grocery Store

The history of point of sale already has been told from a long time ago. The business nowadays is opting for the point-of-sale transactions because it was easy to sell the item to people face to face. This will build some trust between the cashier and the customer. The popular store that always customer need was the grocery store. However, the grocery store nowadays still not has a good feature for the point-of-sale system. The cashier still needs to scan the item through barcode scanner. This idea was led to make the point-of-sale transactions easier in the grocery store by applying the machine learning to the system. The problem for this project is the customer wait for a long time for their point-of-sale transactions to finish when bought the grocery items. The aim of this project is to detect the grocery items with convolutional neural network model for image recognition through camera within the main user interface. The Agile Development Life Cycle (ADLC) method is used in the development of Point-of-Sale System using Machine Learning for Image Recognition in Grocery Store. Moreover, this project is to evaluate the usability of the system using Post-Study System Usability Questionnaire (PSSUQ) approach. The PSSUQ evaluation is evaluated by the users of the system. The results of PSSUQ stated that the users satisfied with the system. The future research for this project is to make the point-of-sale system with a better model in the future. In conclusion, the system is works well and machine learning image recognition model also can detect the grocery item clearly.

Recognition of edible mushrooms based on different image classification methods

The mushroom is one of the most common foods in nature, but there are many different types and many of them are deadly. The aim of this paper is to use computer vision methods to be able to provide a range of assistance to people on safari in identifying mushrooms and the future development of mushroom poisoning and non-poisoning identification. Considering the superior performance of convolutional neural networks in image recognition and classification, the advantages of multi-layer perceptrons and deep structure make it possible to achieve high recognition accuracy, three different CNN network models, GoogLeNet ResNet and SENet were chosen. They have been used to differentiate between toxic and non-poisonous mushrooms on 11,000 mushrooms. The tests showed that SENet had the greatest results, with a recognition accuracy of over 90%. With a recognition accuracy of around 65% each, GoogLeNet and ResNet placed second and third, respectively.

Application of the grade selection of X-ray events using machine learning for a CubeSat mission

X-ray observation covering a wide field of view with high sensitivity is essential in searching for an electromagnetic counterpart of gravitational wave events. A lobster-eye optics (LEO) and a large area CMOS sensor are effective instruments to achieve this goal. Furthermore, thanks to the light weight of LEO, it can be installed on a small platform such as a CubeSat. However, the real-time identification of X-ray events is challenging with restricted resources on space. Therefore, we trained a image recognition network utilizing one of the machine learning models of convolutional neural network (CNN). Then, we use this network to identify X-ray events in the image taken from a CMOS sensor. Moreover, we use a Sony single-board computer, Spresense, that provides ultra-low power consumption and supports machine learning libraries for the process. This paper introduces our machine learning-based X-ray event selection process that is targeted for use on a CubeSat.

Measuring the binary thickness of buccal bone of anterior maxilla in low-resolution cone-beam computed tomography via a bilinear convolutional neural network.

The thickness of the buccal bone of the anterior maxilla is an important aesthetic-determining factor for dental implant, which is divided into the thick (≥1 mm) and thin type (<1 mm). However, as a micro-scale structure that is evaluated through low-resolution cone-beam computed tomography (CBCT), its thickness measurement is error-prone under the circumstance of enormous patients and relatively inexperienced primary dentists. Further, the challenges of deep learning-based analysis of the binary thickness of buccal bone include the substantial real-world variance caused by pixel error, the extraction of fine-grained features, and burdensome annotations. This study built bilinear convolutional neural network (BCNN) with 2 convolutional neural network (CNN) backbones and a bilinear pooling module to predict the binary thickness of buccal bone (thick or thin) of the anterior maxilla in an end-to-end manner. The methods of 5-fold cross-validation and model ensemble were adopted at the training and testing stages. The visualization methods of Gradient Weighted Class Activation Mapping (Grad-CAM), Guided Grad-CAM, and layer-wise relevance propagation (LRP) were used for revealing the important features on which the model focused. The performance metrics and efficacy were compared between BCNN, dentists of different clinical experience (i.e., dental student, junior dentist, and senior dentist), and the fusion of BCNN and dentists to investigate the clinical feasibility of BCNN. Based on the dataset of 4,000 CBCT images from 1,000 patients (aged 36.15±13.09 years), the BCNN with visual geometry group (VGG)16 backbone achieved an accuracy of 0.870 [95% confidence interval (CI): 0.838-0.902] and an area under the receiver operating characteristic (ROC) curve (AUC) of 0.924 (95% CI: 0.896-0.948). Compared with the conventional CNNs, BCNN precisely located the buccal bone wall over irrelevant regions. The BCNN generally outperformed the expert-level dentists. The clinical diagnostic performance of the dentists was improved with the assistance of BCNN. The application of BCNN to the quantitative analysis of binary buccal bone thickness validated the model's excellent ability of subtle feature extraction and achieved expert-level performance. This work signals the potential of fine-grained image recognition networks to the precise quantitative analysis of micro-scale structures.

Deep Residual Networks for Image Recognition

Training deep neural networks is more challenging. We introduce a residual learning framework to simplify the training process for networks that are significantly deeper than those previously used. We specifically redesign the layers to learn residual functions based on the layer inputs, rather than learning functions without any reference. We present thorough empirical evidence indicating that residual networks are easier to optimize and can achieve higher accuracy with significantly increased depth. On the ImageNet dataset, we tested residual networks up to 152 layers deep, which are 8 times deeper than VGG networks but with lower complexity. An ensemble of these networks achieved a 3.57% error rate on the ImageNet test set, securing 1st place in the ILSVRC 2015 classification task. We also analyze networks with 100 and 1000 layers on CIFAR-10. The depth of representations is crucial for various visual recognition tasks. By utilizing extremely deep representations, we achieved a 28% relative improvement on the COCO object detection dataset. Deep residual networks were the basis of our submissions to the ILSVRC and COCO 2015 competitions, where we secured 1st place in ImageNet detection, ImageNet localization, COCO detection, and COCO segmentation tasks.

High performance resistive random access memory based on Ag/TiO2 Nanorods/FTO for image recognition applications

This study investigates the electrical and synaptic properties of Ag/TiO2 nanorod/FTO-based resistive random access memory (RRAM) devices, focusing on the impact of different seed layer thicknesses on nanorod thickness and RRAM performance. The devices exhibit notable achievements, including a high DC endurance of up to 150,000 cycles, a self-compliance function of 10−2 A, and a resistance switching ratio exceeding 100-fold. Notably, the device with a 3.5 μm TiO2 nanorod thickness demonstrates exceptional stability. Analysis of the IV curve reveals that the switching mechanism is attributed to space-charge-limited conduction (SCLC) resulting from electron trapping in oxygen vacancy traps. Furthermore, the device maintains stable synaptic properties even after undergoing 20 cycles of long-term potentiation and depression. When employed in the context of artificial neural networks for image recognition, these devices attain high accuracy rates in recognizing handwritten digits, achieving approximately 90% accuracy for 8 × 8 pixel images and approximately 88% for 28 × 28 pixel images, demonstrating their suitability for practical applications.

Encoding Time Series as Images: A Robust and Transferable Framework for Power System DIM Identification Combining Rules and VGGNet

Digital simulation is very important for the safe and economic operation of power systems. Power system dominant instability mode (DIM) identification is an intractable problem in large-scale transient simulation analysis, because transient instability and short-term voltage instability are often intertwined. Deep learning is a promising way to achieve accurate DIM identification. By encoding time series as images, this paper proposes a robust and transferable DIM identification framework combining curve filtering rules and a dual-channel VGGNet (DCVGG) image recognition network. The curve filtering rules can effectively reduce the redundancy of massive time-domain features, the model parameters, and training time. A skillfully designed method is proposed to encode time series as images and improve the feature extraction capability of convolutional neural networks. The proposed DCVGG network is capable of processing rotor angle and voltage images simultaneously and has the potential to become a pre-trained model, making transfer learning between different power systems possible. Case studies conducted on 8-machine 36-bus system and Northeast China Power Grid demonstrate the proposed framework has better performance, scalability, robustness, and transferable ability than contrastive machine learning methods.

THU260 Development Of Plantar Thermography For Screening Of Diabetic Neuropathy

Abstract Disclosure: V. Panamonta: None. R. Jerawatana: None. P. Ariyaprayoon: None. P. Looareesuwan: None. B. Ongphiphadhanakul: None. C. Sriphrapradang: None. B. Ongphiphadhanakul: None. Background: Diabetic foot ulcers are a major complication of diabetes. They are also the leading cause of nontraumatic lower extremity amputation. The incidence of diabetic foot ulcers is highest in those patients with poor glycemic control and neuropathy. In people with diabetic neuropathy, the feet temperature may change due to decreased blood flow and nerve damage. Thermography has been proposed as a noninvasive modality to identify patients at risk for diabetic foot ulcers. In this study, we used thermography and deep learning to stratify patients with diabetes at risk for developing a foot ulcer. Methods: We prospectively record clinical data and plantar thermogram in adult diabetic patients underwent diabetic foot screening at outpatient clinics at Ramathibodi hospital, Bangkok, Thailand during September to December 2022. Altogether, there were 245 thermal images were analyzed using a deep learning algorithm to determine the risk for diabetic foot ulcers by transfer learning using the pre-trained VGG16 model which is a convolutional neural network for image recognition. Twenty percent of the images were set aside for testing purposes while 20% of the images served as the validation set during training of the neural network. The neural network was trained and weighed more toward higher sensitivity of identifying at risk feet for screening purpose. Results: The study sample consisted of individuals with a mean age of 62.5±12.5 years, 57.1% who had diabetes for more than 10 years. The majority of the participants were females (57.1%). The average body mass index (BMI) was 27.8±9.6 (kg/m2). There were 186 thermal images classified as category 0 and 59 images classified as categories 1 to 3 according to the American Diabetes Association risk classification. The trained neural network achieved 75 % accuracy with 95 % sensitivity and 67 % specificity for classifying thermograms as higher than normal risk in the training dataset. For the testing dataset, the sensitivity and specificity was 75% and 41%, respectively. The accuracy, however, decreased to 50% due mainly to the misclassification of normal thermograms from abnormal. Conclusions: These results suggest that thermography combined with deep learning could potentially be developed for screening purpose to stratify patients at risk of developing diabetic foot ulcers. Further research is needed to validate the results in larger datasets and to explore alternative algorithms for predicting diabetic foot ulcers. Presentation: Thursday, June 15, 2023

Fall Detection Method Based on Convolutional Neural Network

With the improvement of people's living standards and social medical conditions, the average life expectancy is gradually lengthening, and the proportion of the elderly in the global population is also growing. China is a big country with a large population, and the problem of population aging is becoming increasingly prominent and severe. When elderly people live alone at home without others to care for them, falls become the most common and dangerous phenomenon due to the decline of physical function and the influence of certain diseases. Therefore, rapid, efficient and accurate identification and judgment of human falls are of great significance, which can effectively alleviate the threat to the life and health of the elderly and the social medical burden of falls. This paper mainly carried out a fall detection method based on convolutional neural networks, aiming to improve the accuracy of human fall detection through the natural advantages of convolutional neural networks in image recognition, and provide better solutions for medical monitoring and elderly care.