Train Deep Learning Algorithms Research Articles

Image-to-Image Translation-Based Deep Learning Application for Object Identification in Industrial Robot Systems

Industry 4.0 has become one of the most dominant research areas in industrial science today. Many industrial machinery units do not have modern standards that allow for the use of image analysis techniques in their commissioning. Intelligent material handling, sorting, and object recognition are not possible with the machinery we have. We therefore propose a novel deep learning approach for existing robotic devices that can be applied to future robots without modification. In the implementation, 3D CAD models of the PCB relay modules to be recognized are also designed for the implantation machine. Alternatively, we developed and manufactured parts for the assembly of aluminum profiles using FDM 3D printing technology, specifically for sorting purposes. We also apply deep learning algorithms based on the 3D CAD models to generate a dataset of objects for categorization using CGI rendering. We generate two datasets and apply image-to-image translation techniques to train deep learning algorithms. The synthesis achieved sufficient information content and quality in the synthesized images to train deep learning algorithms efficiently with them. As a result, we propose a dataset translation method that is suitable for situations in which regenerating the original dataset can be challenging. The results obtained are analyzed and evaluated for the dataset.

Integration and performance analysis of artificial intelligence and computer vision based on deep learning algorithms

This paper focuses on the analysis of the application effectiveness of the integration of deep learning and computer vision technologies. Deep learning achieves a historic breakthrough by constructing hierarchical neural networks, enabling end-to-end feature learning and semantic understanding of images. The successful experiences in the field of computer vision provide strong support for training deep learning algorithms. The tight integration of these two fields has given rise to a new generation of advanced computer vision systems, significantly surpassing traditional methods in tasks such as machine vision image classification and object detection. In this paper, typical image classification cases are combined to analyze the superior performance of deep neural network models while also pointing out their limitations in generalization and interpretability, proposing directions for future improvements. Overall, the efficient integration and development trend of deep learning with massive visual data will continue to drive technological breakthroughs and application expansion in the field of computer vision, making it possible to build truly intelligent machine vision systems. This deepening fusion paradigm will powerfully promote unprecedented tasks and functions in computer vision, providing stronger development momentum for related disciplines and industries.

BraNet: a mobil application for breast image classification based on deep learning algorithms

Mobile health apps are widely used for breast cancer detection using artificial intelligence algorithms, providing radiologists with second opinions and reducing false diagnoses. This study aims to develop an open-source mobile app named “BraNet” for 2D breast imaging segmentation and classification using deep learning algorithms. During the phase off-line, an SNGAN model was previously trained for synthetic image generation, and subsequently, these images were used to pre-trained SAM and ResNet18 segmentation and classification models. During phase online, the BraNet app was developed using the react native framework, offering a modular deep-learning pipeline for mammography (DM) and ultrasound (US) breast imaging classification. This application operates on a client–server architecture and was implemented in Python for iOS and Android devices. Then, two diagnostic radiologists were given a reading test of 290 total original RoI images to assign the perceived breast tissue type. The reader’s agreement was assessed using the kappa coefficient. The BraNet App Mobil exhibited the highest accuracy in benign and malignant US images (94.7%/93.6%) classification compared to DM during training I (80.9%/76.9%) and training II (73.7/72.3%). The information contrasts with radiological experts’ accuracy, with DM classification being 29%, concerning US 70% for both readers, because they achieved a higher accuracy in US ROI classification than DM images. The kappa value indicates a fair agreement (0.3) for DM images and moderate agreement (0.4) for US images in both readers. It means that not only the amount of data is essential in training deep learning algorithms. Also, it is vital to consider the variety of abnormalities, especially in the mammography data, where several BI-RADS categories are present (microcalcifications, nodules, mass, asymmetry, and dense breasts) and can affect the API accuracy model.Graphical abstract

Application of extended depth of field 3D imagery to tackle the challenges of cryptic species: a use case in the genus Betiscoides Sjöstedt, 1924 (Orthoptera, Caelifera, Lentulidae) and its taxonomic implications

Discovering and handling cryptic diversity among species challenges taxonomists around the world. This is particularly true for the most diverse animal class – the insects – with cryptic diversity, apart from vast species numbers, being one of the main factors that hamper the description of new species. The biodiversity hotspot Cape Floristic Region of South Africa harbors many endemic and yet undescribed insect species, inter alia, Orthoptera. In this study, extended depth of field and 3D imagery enabled for a novel assessment of the external morphological characteristics used for defining and describing the genetically highly diverse genus Betiscoides Sjösdtedt, 1924, leading to a new definition of the genus’ characteristics as well as a revision of character traits of the known species. Two new species are described and a key to all five recognized Betiscoides species is provided. Application standards are derived to enable replicable and reliable image acquisition and measuring. These findings shall contribute to promote efforts being made to establish image based taxonomic identification for researchers worldwide. High-resolution images provide the basis to train deep learning algorithms/ tools, to detect the smallest differences in highly morphologically alike species, and to implement this knowledge in global species monitoring and conservation action to prevent further species loss.

Novel cloud-AIoT fault diagnosis for industrial diesel generators based hybrid deep learning CNN-BGRU algorithm

The industrial diesel generator is essential to provide electric power during emergencies and requires accurate failure diagnosis in preventative maintenance services. This research presents a novel AIoT system-based hybrid deep learning algorithm for online monitoring of the anomaly conditions of 125 kW/250 kW industrial generators. Firstly, the IoT modules with different precise sensors are developed to be installed on 125 kW and 250 kW industrial generators to simulate other faults conditions in the laboratory. Then, the supervised learning operations are deployed to train deep learning algorithms with collected labeled data. The proposed algorithm, a hybrid convolution neural network bidirectional gate recurrent unit (CNN-BGRU), is deployed to extract deep features from long sequential historical signals and classify the anomaly conditions. The experiments in different 10 s, 20 s, and 30 s sampling frequencies are performed to evaluate and analyze with various benchmarks, which demonstrate efficiencies of the proposed CNN-BGRU method with other traditional deep learning algorithms, including RNN, CNN, GRU, LSTM, BRNN, and BGRU methods. The proposed hybrid method accomplishes the supreme augmentation of 29.43 % loss, 36.55 % validating loss, 25.86 % MAE, 27.00 % validating MAE, 16.78 % MSE, 20.64 % validating MSE, 6.36 % Pre, 25.34 % validating Pre, 17.30 % Rec, and 26.91 % validating Rec compared with state-of-the-art approaches. The AIoT system-based hybrid CNN-BGRU algorithm provides better improvement and higher accuracy in fault diagnosis of firefighting pumps for maintenance service in Industry 4.0.

Simulating the cellular context in synthetic datasets for cryo-electron tomography.

Cryo-electron tomography (cryo-ET) allows to visualize the cellular context at macromolecular level. To date, the impossibility of obtaining a reliable ground truth is limiting the application of deep learning-based image processing algorithms in this field. As a consequence, there is a growing demand of realistic synthetic datasets for training deep learning algorithms. In addition, besides assisting the acquisition and interpretation of experimental data, synthetic tomograms are used as reference models for cellular organization analysis from cellular tomograms. Current simulators in cryo-ET focus on reproducing distortions from image acquisition and tomogram reconstruction, however, they can not generate many of the low order features present in cellular tomograms. Here we propose several geometric and organization models to simulate low order cellular structures imaged by cryo-ET. Specifically, clusters of any known cytosolic or membrane bound macromolecules, membranes with different geometries as well as different filamentous structures such as microtubules or actin-like networks. Moreover, we use parametrizable stochastic models to generate a high diversity of geometries and organizations to simulate representative and generalized datasets, including very crowded environments like those observed in native cells. These models have been implemented in a multiplatform open-source Python package, including scripts to generate cryo-tomograms with adjustable sizes and resolutions. In addition, these scripts provide also distortion-free density maps besides the ground truth in different file formats for efficient access and advanced visualization. We show that such a realistic synthetic dataset can be readily used to train generalizable deep learning algorithms.

Generating images of the M87* black hole using GANs

ABSTRACT In this paper, we introduce a novel data augmentation methodology based on Conditional Progressive Generative Adversarial Networks (CPGAN) to generate diverse black hole (BH) images, accounting for variations in spin and electron temperature prescriptions. These generated images are valuable resources for training deep learning algorithms to accurately estimate black hole parameters from observational data. Our model can generate BH images for any spin value within the range of [−1, 1], given an electron temperature distribution. To validate the effectiveness of our approach, we employ a convolutional neural network to predict the BH spin using both the GRMHD images and the images generated by our proposed model. Our results demonstrate a significant performance improvement when training is conducted with the augmented data set while testing is performed using GRMHD simulated data, as indicated by the high R2 score. Consequently, we propose that GANs can be employed as cost-effective models for black hole image generation and reliably augment training data sets for other parametrization algorithms.

DOES - A multimodal dataset for supervised and unsupervised analysis of steel scrap

DOES - Dataset of European scrap classes. Today, scrap is already an important raw material for industry. Due to the transformation to green steel, the secondary raw material scrap will become increasingly important in the coming years. With DOES a free dataset is presented, which represents common non-alloyed European scrap classes. Two important points were considered in this dataset. First, scrap oxidizes under normal external conditions and the visual appearance changes, which plays an important role in visual inspections. Therefore, DOES includes scrap images of different degrees of corrosion attack. Second, images of scrap metal (mostly scrap piles) usually have no intrinsic order. For this reason, a technique to extract many overlapping rectangles from raw images was used, which can be used to train deep learning algorithms without any disadvantage. This dataset is very suitable to develop industrial applications or to research classification algorithms. The dataset was validated by experts and through machine learning models.

Efficient deep learning-based prediction of acoustic parameters in atrium spaces using BIM files

Acoustic simulation tools, although common in the acoustic evaluation of buildings with large atrium space, are costly both in terms of time and resources. This study aims to develop an efficient prediction tool for Reverberation Time (RT) and Sound Pressure Level (SPL) within the atriums using deep learning, with Building Information Modelling (BIM) files as the only input. Initially, a 3-D acoustic simulation model was benchmarked against experimental measurements of RT for an existing building’s atrium, demonstrating an agreement within 15%. Subsequently, BIM files of 60 buildings were used to simulate RT and SPL at various listener locations within their atrium spaces. The BIM files provided essential geometry information, including atrium shape, dimensions, door placements, floor plans, material properties, and sound sources. This dataset was then utilized to train deep learning algorithms, enabling rapid and convenient predictions of RT and SPL for any new building's atrium space. The proposed prediction model serves as a valuable tool for architectural planning and noise regulation during the early design stages. By relying solely on basic building information from the BIM file, this tool obviates the need for time-consuming and computationally expensive simulation software typically used for acoustic evaluations in large atrium spaces.

Deep learning for source localization and seabed classification under dynamic oceanographic environments using explosive sound signals

A deep-learning approach is proposed for simultaneous source localization and seabed classification of underwater signals generated by SUS (Signal, Underwater Sound) MK64 explosives. The charges were deployed in two separate shallow water acoustic experiments on the New England Mudpatch region in 2017 (SBCEX17) and 2022 (SBCEX22) conducted under different oceanographic conditions. The proposed approach uses a multitask learning (MTL) methodology, and convolutional neural networks (CNNs) trained on the spectrogram representation of SUS signals in the 10–199 Hz frequency band where the time-frequency modal dispersion curves exhibit distinct patterns. CNNs were trained on simulated data generated using the normal modes model, ORCA. This synthetic dataset is composed of several sound speed profiles measured in the water column in both experiments and representative seabeds ranging from muddy to sandy sediments. CNNs were tested on the signals measured in the two experiments where the source localization achieved a low mean squared error (0.35–0.8 km), and the seabed classification closely matched existing inversion results in the Mudpatch area. Notably, our method effectively handled oceanographic variations, showcasing the utilization of modal dispersion for training deep learning algorithms. [Work supported by ONR Grant N00014-21-1-2760.]

Artificial intelligence for natural product drug discovery.

Developments in computational omics technologies have provided new means to access the hidden diversity of natural products, unearthing new potential for drug discovery. In parallel, artificial intelligence approaches such as machine learning have led to exciting developments in the computational drug design field, facilitating biological activity prediction and de novo drug design for molecular targets of interest. Here, we describe current and future synergies between these developments to effectively identify drug candidates from the plethora of molecules produced by nature. We also discuss how to address key challenges in realizing the potential of these synergies, such as the need for high-quality datasets to train deep learning algorithms and appropriate strategies for algorithm validation.

EFFECT OF DATA QUALITY ON WATER BODY SEGMENTATION WITH DEEPLABV3+ ALGORITHM

Abstract. Training Deep Learning (DL) algorithms for segmenting features require hundreds to thousands of input data and corresponding labels. Generating thousands of input images and labels requires considerable resources and time. Hence, it is common practice to use opensource imagery data and labels available online. Most of these open-source data have little or no metadata describing their quality or suitability making it problematic for training or evaluating DL models. This study evaluated the effect of data quality on training DeepLabV3+, using Sentinel 2 A/B RGB images and labels obtained from Kaggle. We generated subsets of 256 × 256 pixels, and 10% of these images (802) were set aside for testing. First, we trained and validated the DeepLabV3+ model with the remaining images. Second, we removed images with incorrect labels and trained another DeepLabV3+ network. Finally, we trained the third DeepLabV3+ network after removing images with turbid water or with floating vegetation. All three trained models were evaluated with test images and then we calculated accuracy metrics. As the quality of the input images improved, accuracy of the predicted masks generated from the first model increased from 92.8% to 94.3% in the second model. The third model’s accuracy was 96.4%, demonstrating the network’s ability to better learn and predict water bodies when the input data had fewer class variations. Based on the results we recommend assessing the quality of open-source data for incorrect labels and variations in the target class prior to training DeepLabV3+ or any other DL network.

An approach to the diagnosis of lumbar disc herniation using deep learning models.

Background: In magnetic resonance imaging (MRI), lumbar disc herniation (LDH) detection is challenging due to the various shapes, sizes, angles, and regions associated with bulges, protrusions, extrusions, and sequestrations. Lumbar abnormalities in MRI can be detected automatically by using deep learning methods. As deep learning models gain recognition, they may assist in diagnosing LDH with MRI images and provide initial interpretation in clinical settings. YOU ONLY LOOK ONCE (YOLO) model series are often used to train deep learning algorithms for real-time biomedical image detection and prediction. This study aims to confirm which YOLO models (YOLOv5, YOLOv6, and YOLOv7) perform well in detecting LDH in different regions of the lumbar intervertebral disc. Materials and methods: The methodology involves several steps, including converting DICOM images to JPEG, reviewing and selecting MRI slices for labeling and augmentation using ROBOFLOW, and constructing YOLOv5x, YOLOv6, and YOLOv7 models based on the dataset. The training dataset was combined with the radiologist's labeling and annotation, and then the deep learning models were trained using the training/validation dataset. Results: Our result showed that the 550-dataset with augmentation (AUG) or without augmentation (non-AUG) in YOLOv5x generates satisfactory training performance in LDH detection. The AUG dataset overall performance provides slightly higher accuracy than the non-AUG. YOLOv5x showed the highest performance with 89.30% mAP compared to YOLOv6, and YOLOv7. Also, YOLOv5x in non-AUG dataset showed the balance LDH region detections in L2-L3, L3-L4, L4-L5, and L5-S1 with above 90%. And this illustrates the competitiveness of using non-AUG dataset to detect LDH. Conclusion: Using YOLOv5x and the 550 augmented dataset, LDH can be detected with promising both in non-AUG and AUG dataset. By utilizing the most appropriate YOLO model, clinicians have a greater chance of diagnosing LDH early and preventing adverse effects for their patients.

Speech emotion recognition using the novel PEmoNet (Parallel Emotion Network)

Emotions are very crucial for humans for expressing perception and daily activities such as communication, learning, and decision-making. Human emotion recognition using machines is a very complex task. Recently deep learning techniques have been widely used to automate this task by providing machines with a huge learning capability. However, Speech emotion recognition (SER) is challenging due to language, regional, gender, age, and cultural variations. Most of the previous SER techniques have used only one type of feature representation to train deep learning algorithms, which limits the performance of SER. This paper presents a novel Parallel Emotion Network (PEmoNet) that includes Deep Convolution Neural Network (DCNN) with three parallel arms to address effective SER. The three parallel arms of the proposed PEmoNet accept the Multitaper Mel Frequency Spectrogram (MTMFS), Gammatonegram spectrogram (GS), and Constant Q-Transform Spectrogram (CQTS) as input to improve the feature distinctiveness of the emotion signal. The performance of the proposed SER scheme is evaluated on EMODB and RAVDESS datasets based on accuracy, recall, precision, and F1-score. The proposed technique shows 97.14% and 97.41% accuracy for the EMODB and RAVDESS datasets. It shows that the proposed PEmoNet with different spectral representation inputs helps improve the emotions' distinctiveness and outperforms the existing state of the arts.

Collaborative Privacy-preserving Approaches for Distributed Deep Learning Using Multi-Institutional Data.

Deep learning (DL) algorithms have shown remarkable potential in automating various tasks in medical imaging and radiologic reporting. However, models trained on low quantities of data or only using data from a single institution often are not generalizable to other institutions, which may have different patient demographics or data acquisition characteristics. Therefore, training DL algorithms using data from multiple institutions is crucial to improving the robustness and generalizability of clinically useful DL models. In the context of medical data, simply pooling data from each institution to a central location to train a model poses several issues such as increased risk to patient privacy, increased costs for data storage and transfer, and regulatory challenges. These challenges of centrally hosting data have motivated the development of distributed machine learning techniques and frameworks for collaborative learning that facilitate the training of DL models without the need to explicitly share private medical data. The authors describe several popular methods for collaborative training and review the main considerations for deploying these models. They also highlight publicly available software frameworks for federated learning and showcase several real-world examples of collaborative learning. The authors conclude by discussing some key challenges and future research directions for distributed DL. They aim to introduce clinicians to the benefits, limitations, and risks of using distributed DL for the development of medical artificial intelligence algorithms. ©RSNA, 2023 Quiz questions for this article are available in the supplemental material.

A Bayesian Optimization-Based LSTM Model for Wind Power Forecasting in the Adama District, Ethiopia

Renewable energies, such as solar and wind power, have become promising sources of energy to address the increase in greenhouse gases caused by the use of fossil fuels and to resolve the current energy crisis. Integrating wind energy into a large-scale electric grid presents a significant challenge due to the high intermittency and nonlinear behavior of wind power. Accurate wind power forecasting is essential for safe and efficient integration into the grid system. Many prediction models have been developed to predict the uncertain and nonlinear time series of wind power, but most neglect the use of Bayesian optimization to optimize the hyperparameters while training deep learning algorithms. The efficiency of grid search strategies decreases as the number of hyperparameters increases, and computation time complexity becomes an issue. This paper presents a robust and optimized long-short term memory network for forecasting wind power generation in the day ahead in the context of Ethiopia’s renewable energy sector. The proposal uses Bayesian optimization to find the best hyperparameter combination in a reasonable computation time. The results indicate that tuning hyperparameters using this metaheuristic prior to building deep learning models significantly improves the predictive performances of the models. The proposed models were evaluated using MAE, RMSE, and MAPE metrics, and outperformed both the baseline models and the optimized gated recurrent unit architecture.

The Distributed Deep Learning Paradigms for Detection of Weeds from Crops in Indian Agricultural Farms

Weeds are a major threat to crops, making early detection critical for maintaining agricultural productivity. Weeds are generally toxic, equipped with thorns and burrs, and can disrupt crop management by contaminating harvests. This research aims to identify weeds in a field using image processing and deep learning techniques. Images were collected from an Indian farm and pre-processed using image processing techniques. The images were then analysed to extract features that distinguish between weed and crop properties. Traditional crop weed identification methods mainly focused on identifying weeds directly but weed species can vary significantly. This study proposes a method that combines deep learning and image processing technology. Identifying weeds in crops is a challenging task that has been addressed through image processing, feature extraction, and image labelling to train deep learning algorithms. The study examines the performance of various deep learning algorithms and convolution neural networks to detect weeds using images obtained from an Indian crop field. Once the input image is identified as a weed or not, the crop class prediction is made. These results could have significant implications for optimizing agricultural fertilizer usage, leading to increased crop yields and less environmental impact.

Smart Attendance Automation System

The smart attendance marking system, which is based on face recognition technology, is a modern solution that may aid in the automation of the attendance management process in a variety of organizations, including schools, colleges, and corporations. Using facial recognition technology, the research team hopes to create an efficient system for tracking and recording attendance. The system will be developed by gathering and analysing data in order to construct a database of faces that will be used to train deep learning algorithms to recognise faces effectively. Compared to traditional attendance tracking systems, the smart attendance marking system based on face recognition technology has various advantages. For starters, it saves time and lowers manual labour because attendance is automatically tracked. Second, it enhances precision, lowering the possibility of errors. Third, it delivers real-time attendance updates, making it simple for instructors or administration to monitor attendance and take appropriate action. The aim of the work is expected to improve attendance management in various organisations by optimising attendance monitoring methods while decreasing manual effort and errors. Because the system may be connected with existing attendance management systems or utilised as a standalone solution, it is a versatile and adaptable option for a variety of organisations. The paper’s effort is to design, develop, and implement an efficient system that can track and record attendance accurately using facial recognition technology, giving benefits such as time savings, accuracy, real-time updates, and better security.

STARC: Deep learning Algorithms’ modelling for STructured analysis of retina classification

Retina is the heart of an eye which generates electrical impulses due to light sensitivity. The vessel formation in human eye is an essential key for diagnosing ophthalmological conditions. This paper aims to diagnose ophthalmological conditions through deep learning models and provide advancements in early detection of ophthalmological conditions for proper treatment to protect patient’s vision, and for health care giver worldwide. STARE dataset is used for this study which consists over 385 retinal images of 14 ophthalmological defects such as BRAO, CRAO, etc. This dataset is further pre-processed over the techniques such as augmentation, normalization, etc for obtaining the best refined features for training deep learning algorithms. This paper broadly implements 5 deep learning algorithms i.e., EfficientNet, 3-Layers CNN, InceptionV2, ResNet-50, VGG-16. These models are trained number of times over tuned hyperparameters such as batch size etc and evaluated over 4 performance metrics over weighted averaged and macro averaged of precision, recall, F1-score, and accuracy to acquire the best performing model. EfficientNet performed the best with 98.43% accuracy, macro averaged f-1 score, recall, precision as 98.37%, 99.16%, 97.91% and weighted averaged f-1 score, recall, precision, as 98.50%, 98.43%, 98.82% over batch size 64. In this study, the performance compared with the existing literature achieved higher accuracy due to the unique model and its configuration, hyperparameter tuning and pre-processing techniques for the 14 classes of retinal defects. The future work includes classifying more ophthalmological conditions, adding more parameters from blood, etc.

DETECTION OF STRUCTURAL CRACKS OF AN AIRCRAFT USING DEEP NEURAL NETWORKS


 
 
 Thanks to developments in machine learning (ML), particularly deep learning, which now has the greatest performance among algorithms, narrow artificial intelligence, often known as "weak AI," has progressed during the past few years. further machine learning for deep learning to provide potent model parameters that can forecast the occurrence of certain events in the future, massive volumes of data, also referred to as "big data” and must be rapidly collected. Such a big damage event dataset is not accessible in many other fields, such as visual inspection of aeroplanes, and this makes it difficult to train deep learning algorithms to work effectively. Good at spotting physical damage to aircraft structures. In order to reach this human-level intelligence in aircraft damage inspection, it is possible to include inductive bias into deep learning. This paper provides an illustration of how to incorporate expertise in aircraft engineering into the creation of deep learning algorithms. The effectiveness of our method, which builds a deep convolutional neural network that categorises crack lengths based on break propagation curves acquired from fatigue testing, was shown on aerospace grade aluminium samples.