Deep Convolutional Neural Network Model Research Articles

Classification of the L-, H-mode, and plasma-free state: Convolutional neural networks and variational autoencoders on the edge reflectometer for KSTAR.

Classifying and monitoring the L-, H-mode, and plasma-free state are essential for the stable operational control of tokamaks. Edge reflectometry measures plasma density profiles, but the large volume of data and complexity in reconstruction pose significant challenges. There is a need for efficient methods to analyze complex reflectometer data in real-time, which can be addressed using advanced computational techniques. Here, we show that machine learning (ML) techniques can classify discharge states using raw signal data from an edge reflectometer installed on the Korea Superconducting Tokamak Advanced Research. The deep convolutional neural network models achieved classification accuracy of up to 99% when using 2D spectrogram inputs, demonstrating a significant improvement over 1D raw signal inputs. Additionally, the variational autoencoder model effectively clustered the discharge states in the latent space without any label information, further validating the model's capability to classify discharge states. These results suggest that the ML model can effectively handle the complexity of reflectometer data and accurately classify plasma discharge states. This approach not only facilitates real-time diagnosis but also reduces the need for manual data processing.

Active Learning for Deep Visual Tracking.

Convolutional neural networks (CNNs) have been successfully applied to the single target tracking task in recent years. Generally, training a deep CNN model requires numerous labeled training samples, and the number and quality of these samples directly affect the representational capability of the trained model. However, this approach is restrictive in practice, because manually labeling such a large number of training samples is time-consuming and prohibitively expensive. In this article, we propose an active learning method for deep visual tracking, which selects and annotates the unlabeled samples to train the deep CNN model. Under the guidance of active learning, the tracker based on the trained deep CNN model can achieve competitive tracking performance while reducing the labeling cost. More specifically, to ensure the diversity of selected samples, we propose an active learning method based on multiframe collaboration to select those training samples that should be and need to be annotated. Meanwhile, considering the representativeness of these selected samples, we adopt a nearest-neighbor discrimination method based on the average nearest-neighbor distance to screen isolated samples and low-quality samples. Therefore, the training samples' subset selected based on our method requires only a given budget to maintain the diversity and representativeness of the entire sample set. Furthermore, we adopt a Tversky loss to improve the bounding box estimation of our tracker, which can ensure that the tracker achieves more accurate target states. Extensive experimental results confirm that our active-learning-based tracker (ALT) achieves competitive tracking accuracy and speed compared with state-of-the-art trackers on the seven most challenging evaluation benchmarks. Project website: https://sites.google.com/view/altrack/.

Development of fig fruit ripeness classification using convolutional neural network

This study presents the design and evaluation of a deep convolutional neural network (CNN) model for accurately classifying fig ripeness stages. Traditionally, fruit ripeness classification has been conducted manually, which presents several drawbacks, including heavy reliance on human labor and inconsistencies in determining fruit ripeness. By leveraging advanced deep learning techniques, specifically CNNs, this research aims to automate the fig ripeness classification process. The CNN architecture was developed and trained using MATLAB software, targeting three ripeness categories: ripe, half-ripe, and unripe. The methodology involved pre-processing the fig images and configuring the CNN model with multiple convolutional, batch normalization, and max pooling layers specifically for fig classification tasks. The final CNN model achieved an impressive accuracy rate of 94.44%, significantly surpassing results from previously reported studies. The developed model is a promising tool for automating fig ripeness classification, contributing to advancements in precision agriculture and smart farming technologies.

Development and application of an artificial intelligence-assisted endoscopy system for diagnosis of Helicobacter pylori infection: a multicenter randomized controlled study

BackgroundThe early diagnosis and treatment of Heliobacter pylori (H.pylori) gastrointestinal infection provide significant benefits to patients. We constructed a convolutional neural network (CNN) model based on an endoscopic system to diagnose H. pylori infection, and then examined the potential benefit of this model to endoscopists in their diagnosis of H. pylori infection.Materials and methodsA CNN neural network system for endoscopic diagnosis of H.pylori infection was established by collecting 7377 endoscopic images from 639 patients. The accuracy, sensitivity, and specificity were determined. Then, a randomized controlled study was used to compare the accuracy of diagnosis of H. pylori infection by endoscopists who were assisted or unassisted by this CNN model.ResultsThe deep CNN model for diagnosis of H. pylori infection had an accuracy of 89.6%, a sensitivity of 90.9%, and a specificity of 88.9%. Relative to the group of endoscopists unassisted by AI, the AI-assisted group had better accuracy (92.8% [194/209; 95%CI: 89.3%, 96.4%] vs. 75.6% [158/209; 95%CI: 69.7%, 81.5%]), sensitivity (91.8% [67/73; 95%CI: 85.3%, 98.2%] vs. 78.6% [44/56; 95%CI: 67.5%, 89.7%]), and specificity (93.4% [127/136; 95%CI: 89.2%, 97.6%] vs. 74.5% [114/153; 95%CI: 67.5%, 81.5%]). All of these differences were statistically significant (P < 0.05).ConclusionOur AI-assisted system for diagnosis of H. pylori infection has significant ability for diagnostic, and can improve the accuracy of endoscopists in gastroscopic diagnosis.Trial registrationThis study was approved by the Ethics Committee of Daping Hospital (10/07/2020) (No.89,2020) and was registered with the Chinese Clinical Trial Registration Center (02/09/2020) (www.chictr.org.cn; registration number: ChiCTR2000037801).

Spoofing Detection in Images Using Eques Chimp Optimization-based Deep Convolutional Neural Network Model

Spoofing Detection in Images Using Eques Chimp Optimization-based Deep Convolutional Neural Network Model

Deep Learning Methods for Chest X-Ray Imaging-Based COVID-19 Pneumonia Detection

The COVID-19 pandemic currently underway has highlighted a need for fast and accurate screening tools to help identify the disease, particularly in resource-poor settings. In this paper, a deep convolutional neural network (CNN) model is proposed for the automatic detection of COVID-19 pneumonia using chest X-ray images that helps in reducing the cost and processing time compared to traditional testing procedures. It uses several deep-structured medical image data types and improvement of the deep learning model architecture to achieve good diagnostic accuracy. Our hybrid CNN architecture integrates VGG-16 for feature extraction and ResNet-50 for pattern complexity assessment in detecting fine changes characteristic of COVID-19 pneumonia. Our dataset for this study is a collection of few thousands labelled chest X-ray images categorized in three classes, which are COVID-19, viral pneumonia and healthy cases. More advanced data preprocessing methods such as normalized, augmentation, and filtered noise has also been carried out which enhances in the model performance. We have high accuracy, recall, and good F1-score in the experimental results proving model robustness are applicable in real-world clinical scenarios. These research results speak to the importance of AI for improving diagnosis workflows, especially in fast-to-deploy large scale scenarios needed during a pandemic to cope with healthcare burdens.

Computer Vision Identification of Trachomatous Inflammation-Follicular Using Deep Learning.

Trachoma surveys are used to estimate the prevalence of trachomatous inflammation-follicular (TF) to guide mass antibiotic distribution. These surveys currently rely on human graders, introducing a significant resource burden and potential for human error. This study describes the development and evaluation of machine learning models intended to reduce cost and improve reliability of these surveys. Fifty-six thousand seven hundred twenty-five everted eyelid photographs were obtained from 11,358 children of age 0 to 9 years in a single trachoma-endemic region of Ethiopia over a 3-year period. Expert graders reviewed all images from each examination to determine the estimated number of tarsal conjunctival follicles and the degree of trachomatous inflammation-intense. The median estimate of the 3 grader groups was used as the ground truth to train a MobileNetV3 large deep convolutional neural network to detect cases with TF. The classification model predicted a TF prevalence of 32%, which was not significantly different from the human consensus estimate (30%; 95% confidence interval of difference, -2 to +4%). The model had an area under the receiver operating characteristic curve of 0.943, F1 score of 0.923, 88% accuracy, 83% sensitivity, and 91% specificity. The area under the receiver operating characteristic curve increased to 0.995 when interpreting nonborderline cases of TF. Deep convolutional neural network models performed well at classifying TF and detecting the number of follicles evident in conjunctival photographs. Implementation of similar models may enable accurate, efficient, large-scale trachoma screening. Further validation in diverse populations with varying TF prevalence is needed before implementation at scale.

Lung nodule detection using Eyrie Flock-based Deep Convolutional Neural Network

PROBLEM: Lung cancer is a dangerous and deadly disease with high mortality and reduced survival rates. However, the lung nodule diagnosis performance is limited by its heterogeneity in terms of texture, shape, and intensity. Furthermore, the high degree of resemblance between the lung nodules and the tissues that surround the lung nodules makes the building of a reliable detection model more difficult. Moreover, there are several methods for diagnosing and grading lung nodules; still, the accuracy of detection with the variations in intensity is a challenging task. AIM &amp; METHODS: For the detection of lung nodules and grading, this research proposes an Eyrie Flock Optimization-based Deep Convolutional Neural Network (Eyrie Flock-DeepCNN). The proposed Eyrie Flock Optimization integrates the fishing characteristics of Eyrie’s and the flocking characteristics of Tusker to accelerate the convergence speed which inturns enhance the training process and improve the generalization performance of the DeepCNN model. In the Eyrie Flock optimization, two optimal issues are considered: (i) segmenting the lung nodule and (ii) fine-tuning hyperparameters of Deep CNN. RESULTS: The capability of the newly developed method is evaluated by the terms of Specificity, Sensitivity, and Accuracy, attaining 98.96%, 95.21%, and 94.12%, respectively. CONCLUSION: Efficiently utilized the Deep CNN along with the help of the Eyrie Flock optimization algorithm which enhances the efficiency of the classifier and convergence of the model.

E-DFu-Net: An efficient deep convolutional neural network models for Diabetic Foot Ulcer classification.

The Diabetic Foot Ulcer (DFU) is a severe complication that affects approximately 33% of diabetes patients globally, often leading to limb amputation if not detected early. This study introduces an automated approach for identifying and classifying DFU using transfer learning. DFU is typically categorized into ischemic and infection states, which are challenging to distinguish visually. We evaluate the effectiveness of pre-trained Deep Convolutional Neural Network (DCNN) models for autonomous DFU detection. Seven models are compared: EfficientNetB0, DenseNet121, ResNet101, VGG16, MobileNetV2, InceptionV3, and InceptionResNetV2. Additionally, we propose E-DFu-Net, a novel model derived from existing architectures, designed to mitigate overfitting. Experimental results demonstrate that E-DFu-Net achieves remarkable performance, with 97% accuracy in ischemia classification and 92% in infection classification. This advancement enhances current methodologies and aids practitioners in effectively detecting DFU cases.

Prediction of plasma rotation velocity and ion temperature profiles in EAST Tokamak using artificial neural network models

Abstract Artificial neural network models have been developed to predict rotation velocity and ion temperature profiles on the EAST tokamak based on spectral measurements from the X-ray crystal spectrometer. Both Deep Neural Network (DNN) and Convolutional Neural Network (CNN) models have been employed to infer line-integrated ion temperatures. The predicted results from these two models exhibit a strong correlation with the target values, providing an opportunity for cross-validation to enhance prediction accuracy. Notably, the computational speed of these models has been significantly increased, surpassing traditional methods by over tenfold. Furthermore, the investigation of input data range and error prediction serves as the foundation for future automated calculation process. Finally, CNNs have also been employed to predict line-integrated rotation velocity profiles and inverted ion temperature profiles for their robustness in the training process. It is noted that these algorithms are not restricted to any specific physics model and can be readily adapted to various fusion devices.

The use of digital thread for reconstruction of local fiber orientation in a compression molded pin bracket via deep learning

A deep convolutional neural network (DCNN) was used for microstructure reconstruction using artificial intelligence (MR-AI) by predicting local average fiber orientation distributions (FOD) in a 3D prepreg platelet molded composite (PPMC) pin bracket. To train the MR-AI model, surface strain fields from residual stresses simulated in PPMC plates were used as the input to the DCNN. A training dataset included PPMC plates with various degrees of global fiber alignment, based on the information obtained from high-fidelity flow simulation of a pin bracket. The MR-AI model was then deployed to analyze FOD in the 3D pin bracket by conducting thermo-elastic residual stress analysis. Initially, the MR-AI model was established entirely on the synthetic simulation data. Then, a μCT scan of a physically molded pin bracket was used to create a finite element model that provided data for additional validation of the DCNN model. For the μCT scan finite element pin bracket the MR-AI model predicted the distribution of fiber orientation tensor components with MAE of 0.10 indicating a global prediction error of 10 %. For the flow simulated pin bracket, the MR-AI model predicted the distribution of fiber orientation tensor components with a global prediction error of 11 %. The MR-AI model showed the ability to predict regions of varying alignment in the base and flange of the pin bracket. The proposed MR-AI methodology allows for rapid prediction of FOD in geometrically complex parts and offers a promising path to detecting unique fiber orientation states in molded components.

Beehive Smart Detector Device for the Detection of Critical Conditions That Utilize Edge Device Computations and Deep Learning Inferences.

This paper presents a new edge detection process implemented in an embedded IoT device called Bee Smart Detection node to detect catastrophic apiary events. Such events include swarming, queen loss, and the detection of Colony Collapse Disorder (CCD) conditions. Two deep learning sub-processes are used for this purpose. The first uses a fuzzy multi-layered neural network of variable depths called fuzzy-stranded-NN to detect CCD conditions based on temperature and humidity measurements inside the beehive. The second utilizes a deep learning CNN model to detect swarming and queen loss cases based on sound recordings. The proposed processes have been implemented into autonomous Bee Smart Detection IoT devices that transmit their measurements and the detection results to the cloud over Wi-Fi. The BeeSD devices have been tested for easy-to-use functionality, autonomous operation, deep learning model inference accuracy, and inference execution speeds. The author presents the experimental results of the fuzzy-stranded-NN model for detecting critical conditions and deep learning CNN models for detecting swarming and queen loss. From the presented experimental results, the stranded-NN achieved accuracy results up to 95%, while the ResNet-50 model presented accuracy results up to 99% for detecting swarming or queen loss events. The ResNet-18 model is also the fastest inference speed replacement of the ResNet-50 model, achieving up to 93% accuracy results. Finally, cross-comparison of the deep learning models with machine learning ones shows that deep learning models can provide at least 3-5% better accuracy results.

Learning-assisted specklegram analysis for recognition of simultaneous weights on multimode optical fiber

A deep learning-based recognition of multimode fiber (MMF) specklegrams for various simultaneous weights is presented in this work. Five different random locations have been considered along the length of MMF and the specklegram images are recorded corresponding to seven different combinations of random simultaneous weights applied at these locations. A popular deep learning convolutional neural network (CNN) model, VGG-16 is employed on these images for the recognition of these seven weight combinations. The impact of acoustic vibrations, laser power, external temperature, and image sizes on the recognition accuracy is examined. A 100% recognition accuracy is attained and a negligible accuracy variation of ∼1.9% for acoustic vibrations as well as for changing laser power is observed, whereas a drastic fall in accuracy is observed in case of change in image sizes less than 80 × 80 pixels. Also, a negligible variation of ∼2% is observed for the applied external temperature. The heart of our work lies in the accumulation of a diverse, large volume of specklegram dataset by virtue of conducting brute force experiments that take care of eradication of model overfitting. The proposed proof-of-concept scheme might be useful for low-cost, efficacious, self-assisted multi-weight analysis in structural health monitoring.

Real-time thermography for breast cancer detection with deep learning

In this study, we propose a framework that enhances breast cancer classification accuracy by preserving spatial features and leveraging in situ cooling support. The framework utilizes real-time thermography video streaming for early breast cancer detection using Deep Learning models. Inception v3, Inception v4, and a modified Inception Mv4 were developed using MATLAB 2019. However, the thermal camera was connected to a mobile phone to capture images of the breast area for classification of normal and abnormal breast. This study’s training dataset included 1000 thermal images, where a FLIR One Pro thermal camera connected to a mobile device was used for the imaging process. Of the 1000 images obtained, 700 images were considered for the normal breast thermography class while the 300 images were suitable for the abnormal class. We evaluate Deep Convolutional Neural Network models, such as Inception v3, Inception v4, and a modified Inception Mv4. Our results demonstrate that Inception Mv4, with real-time video streaming, efficiently detects even the slightest temperature contrast in breast tissue sequences achieving a 99.748% accuracy in comparison to a 99.712% and 96.8% for Inception v4 and v3, respectively. The use of in situ cooling gel further enhances image acquisition efficiency and detection accuracy. Interestingly, increasing the tumor surface temperature by 0.1% leads to an average 7% improvement in detection and classification accuracy. Our findings support the effectiveness of Inception Mv4 for real-time breast cancer detection, especially when combined with in situ cooling gel and varying tumor temperatures. In conclusion, future research directions should focus on incorporating thermal video clips into the thermal images database, utilizing high-quality thermal cameras, and exploring alternative Deep Learning models for improved breast cancer detection.

A precise grape yield prediction method based on a modified DCNN model

This paper presents an innovative method for predicting grape yield using a modified deep convolutional neural network (M−DCNN). This approach integrates advanced image processing techniques with deep learning algorithms to analyze high-resolution images of vineyards, enabling precise yield quantification. The M−DCNN model incorporates novel convolutional layers and activation functions to enhance accuracy and generalization capabilities. Extensive experiments were conducted using a varied dataset of grape varieties and vineyard conditions. The method involves image processing to identify grape clusters and monitor their growth at different stages. We evaluated the performance of several algorithms, including K-nearest neighbors (KNN), multinomial logistic regression (MLR), support vector machines (SVM), random forests (RF), deep convolutional neural network (DCNN), and M−DCNN, throughout the training and testing phases. The M−DCNN achieved the highest training accuracy at 97.35%, which represents a 0.85% improvement over the standard DCNN. Furthermore, in detecting grape berries, our method attained an average F1 score of 73.32% and coefficients of determination exceeding 0.98, showcasing high accuracy and precision. The results indicate that the M−DCNN model surpasses traditional yield estimation methods and existing machine learning models, affirming its potential to significantly enhance precision agriculture. This method offers a robust tool for accurate grape yield forecasting, facilitating informed decision-making in viticulture.

Speech enhancement using deep complex convolutional neural network (DCCNN) model

Speech enhancement using deep complex convolutional neural network (DCCNN) model

A novel local feature fusion architecture for wind turbine pitch fault diagnosis with redundant feature screening

The safe and reliable operation of the pitch system is essential for the stable and efficient operation of a wind turbine (WT). The pitch fault data collected by supervisory control and data acquisition systems (SCADA) often contain a wide variety of variables, leading to redundant features that interfere with the accuracy of final diagnosis results, making it difficult to meet requirements. Also, the problem of extracting only local features while ignoring global information is present in the feature extraction process using the deep Convolutional Neural Network (CNN) model. To address these issues, the global average correlation coefficient is proposed in this article to measure the correlation between multiple variables in SCADA data. By considering the correlation among multiple variables comprehensively, redundant features are effectively eliminated, enhancing the accuracy of fault diagnosis. Furthermore, a new local amplification fusion architecture network (LAFA-Net) based on multi-head attention (MHA) is introduced. An efficient local feature extraction module, designed to enhance the model’s perception of detailed features while maintaining global context information, is first introduced. LAFA-Net integrates the advantages of CNN and MHA, efficiently extracting and fusing valuable features from filtered data for both local and global aspects. Experiments on real pitch fault data demonstrate that the global average correlation coefficient effectively screens out redundant features in the dataset that negatively impact fault diagnosis results, thereby improving diagnosis efficiency and accuracy. The LAFA-Net model, capable of accurately diagnosing multiple types of pitch faults, shows a superior classification effect and accuracy compared to several advanced models, along with a faster convergence speed.

TEFLON: Thermally Efficient Dataflow-aware 3D NoC for Accelerating CNN Inferencing on Manycore PIM Architectures

Resistive random-access memory (ReRAM)-based processing-in-memory (PIM) architectures are used extensively to accelerate inferencing/training with convolutional neural networks (CNNs). Three-dimensional (3D) integration is an enabling technology to integrate many PIM cores on a single chip. In this work, we propose the design of a t hermally e fficient data flo w-aware monolithic 3D (M3D) N oC architecture referred to as TEFLON to accelerate CNN inferencing without creating any thermal bottlenecks. TEFLON reduces the Energy-Delay-Product (EDP) by 42%, 46%, and 45% on an average compared to a conventional 3D mesh NoC for systems with 36-, 64-, and 100-PIM cores, respectively. TEFLON reduces the peak chip temperature by 25 K and improves the inference accuracy by up to 11% compared to sole performance-optimized SFC-based counterpart for inferencing with diverse deep CNN models using CIFAR-10/100 datasets on a 3D system with 100-PIM cores.

Cosine similarity-guided knowledge distillation for robust object detectors

This paper presents a Cosine Similarity-Based Knowledge Distillation (CSKD) for robust, lightweight object detectors. Knowledge Distillation (KD) has been effective in enhancing the performance of compact models in image classification by leveraging deep CNN models. However, the complex and multifaceted nature of object detection, characterized by its modular design and multitasking requirements, poses significant challenges for traditional KD techniques. These challenges are further compounded by the conventional reliance on the Mean Squared Error (MSE) loss function and the limited application of enhanced feature representations to the training phase. Addressing these limitations, the proposed CSKD method combines cosine similarity guidance with MSE loss to facilitate a more effective knowledge transfer from the teacher model to the student model. This is achieved by distilling both intermediate features and prediction outputs, aided by an assistant prediction branch designed to learn directly from the teacher’s predictions. This dual-faceted distillation strategy enables the student model to better mimic the teacher model’s behavior, leading to improved performance. The proposed method demonstrates versatility and robustness across various object detector architectures without the need for additional feature enhancement layers during training. Notably, employing ResNet-50 as the teacher model and ResNet-18 as the student model, we achieve new benchmarks in KD for object detection across several architectures, including Faster-RCNN, RetinaNet, FCOS, and GFL, with respective mAP scores of 36.6, 35.2, 35.9, and 38.9. These results highlights the effectiveness of CSKD in advancing the state-of-the-art in KD for object detection, offering a compelling solution to the challenges previously faced by traditional KD methods in this domain. The code of the proposed CSKD is available at https://github.com/swkdn16/CSKD.

Human Visual Pathways for Action Recognition Versus Deep Convolutional Neural Networks: Representation Correspondence in Late But Not Early Layers.

Deep convolutional neural networks (DCNNs) have attained human-level performance for object categorization and exhibited representation alignment between network layers and brain regions. Does such representation alignment naturally extend to other visual tasks beyond recognizing objects in static images? In this study, we expanded the exploration to the recognition of human actions from videos and assessed the representation capabilities and alignment of two-stream DCNNs in comparison with brain regions situated along ventral and dorsal pathways. Using decoding analysis and representational similarity analysis, we show that DCNN models do not show hierarchical representation alignment to human brain across visual regions when processing action videos. Instead, later layers of DCNN models demonstrate greater representation similarities to the human visual cortex. These findings were revealed for two display formats: photorealistic avatars with full-body information and simplified stimuli in the point-light display. The discrepancies in representation alignment suggest fundamental differences in how DCNNs and the human brain represent dynamic visual information related to actions.