Fine-tuned VGG16 Research Articles

Preserving authenticity: transfer learning methods for detecting and verifying facial image manipulation

Facial retouching in supporting documents can have adverse effects, undermining the credibility and authenticity of the information presented. This paper presents a comprehensive investigation into the classification of retouched face images using a fine-tuned pre-trained VGG16 model. We explore the impact of different train-test split strategies on the performance of the model and also evaluate the effectiveness of two distinct optimizers. The proposed fine-tuned VGG16 model with “ImageNet” weight achieves a training accuracy of 99.34 % and a validation accuracy of 97.91 % over 30 epochs on the ND-IIITD retouched faces dataset. The VGG16_Adam model gives a maximum classification accuracy of 96.34 % for retouched faces and an overall accuracy of 98.08 %. The experimental results show that the 50 % - 25 % train-test split ratio outperforms other split ratios mentioned in the paper. The demonstrated work shows that using a Transfer Learning approach reduces computational complexity and training time, with a max. training duration of 39.34 min for the proposed model.

End-to-end tire defect detection model based on transfer learning techniques

Visual inspection of defective tires post-production is vital for human safety, as faulty tires can lead to explosions, accidents, and loss of life. With the advancement of technology, transfer learning (TL) plays an influential role in many computer vision applications, including the tire defect detection problem. However, automatic tire defect detection is difficult for two reasons. The first is the presence of complex anisotropic multi-textured rubber layers. Second, there is no standard tire X-ray image dataset to use for defect detection. In this study, a TL-based tire defect detection model is proposed using a new dataset from a global tire company. First, we collected and labeled the dataset consisting of 3366 X-ray images of faulty tires and 20,000 images of qualified tires. Although the dataset covers 15 types of defects arising from different design patterns, our primary focus is on binary classification to detect the presence or absence of defects. This challenging dataset was split into 70, 15, and 15% for training, validation, and testing, respectively. Then, nine common pre-trained models were fine-tuned, trained, and tested on the proposed dataset. These models are Xception, InceptionV3, VGG16, VGG19, ResNet50, ResNet152V2, DenseNet121, InceptionResNetV2, and MobileNetV2. The results show that the fine-tuned VGG19, DenseNet21 and InceptionNet models achieve compatible results with the literature. Moreover, the Xception model outperformed the compared TL models and literature methods in terms of recall, precision, accuracy, and F1 score. Moreover, it achieved on the testing dataset 73.7, 88, 80.2, and 94.75% of recall, precision, F1 score, and accuracy, respectively, and on the validation dataset 73.3, 90.24, 80.9, and 95% of recall, precision, F1 score, and accuracy, respectively.

MRI super-resolution using similarity distance and multi-scale receptive field based feature fusion GAN and pre-trained slice interpolation network

Challenges arise in achieving high-resolution Magnetic Resonance Imaging (MRI) to improve disease diagnosis accuracy due to limitations in hardware, patient discomfort, long acquisition times, and high costs. While Convolutional Neural Networks (CNNs) have shown promising results in MRI super-resolution, they often don't look into the structural similarity and prior information available in consecutive MRI slices. By leveraging information from sequential slices, more robust features can be obtained, potentially leading to higher-quality MRI slices. We propose a multi-slice two-dimensional (2D) MRI super-resolution network that combines a Generative Adversarial Network (GAN) with feature fusion and a pre-trained slice interpolation network to achieve three-dimensional (3D) super-resolution. The proposed model requires consecutively acquired three low-resolution (LR) MRI slices along a specific axis, and achieves the reconstruction of the MRI slices in the remaining two axes. The network effectively enhances both in-plane and out-of-plane resolution along the sagittal axis while addressing computational and memory constraints in 3D super-resolution. The proposed generator has a in-plane and out-of-plane Attention (IOA) network that fuses both in-plane and out-plane features of MRI dynamically. In terms of out-of-plane attention, the network merges features by considering the similarity distance between features and for in-plane attention, the network employs a two-level pyramid structure with varying receptive fields to extract features at different scales, ensuring the inclusion of both global and local features. Subsequently, to achieve 3D MRI super-resolution, a pre-trained slice interpolation network is used that takes two consecutive super-resolved MRI slices to generate a new intermediate slice. To further enhance the network performance and perceptual quality, we introduce a feature up-sampling layer and a feature extraction block with Scaled Exponential Linear Unit (SeLU). Moreover, our super-resolution network incorporates VGG loss from a fine-tuned VGG-19 network to provide additional enhancement. Through experimental evaluations on the IXI dataset and BRATS dataset, using the peak signal-to-noise ratio (PSNR), structural similarity index measure (SSIM) and the number of training parameters, we demonstrate the superior performance of our method compared to the existing techniques. Also, the proposed model can be adapted or modified to achieve super-resolution for both 2D and 3D MRI data.

Development of an Approach for Image Forgery Detection Using Machine Learning Algorithms

Digital picture fraud detection is an increasing societal necessity due to the importance of verified images. The detection of picture copying, splicing, retouching, and re-sampling forgeries is included. In the absence of digital signatures or watermarks, passive picture authentication may serve as an alternative to active authentication. Passive techniques, every so often recognized as blind techniques, could take place without preceding knowledge of the picture or its reference. Identifying counterfeiting picture or tampering was a research field for long a period of time, triggered via the Internet, online platforms, social messaging platforms, and extensive digital image usage. The rate of failure could be a key factor for examining the alteration of picture or forgery, among other existing methods. The research applies almost six common algorithms related to machine learning in order to extract features from Lightweight, Spatial Exploitation, and Residual deep learning models on benchmark datasets MICC-F220, Columbia, and CoMoFoD. The models of incorporated deep learning could consist of AlexNet, GoogleNet, VGG16, VGG19, SqueezeNet, MobileNetV2, ShuffleNet, ResNet-18, ResNet-50, and ResNet-101 for spatial exploitation. Fine-tuning is applied to the top three deep learning models, optimizing hyperparameters centered on indicators of performance for every single benchmark dataset. Tweaked SqueezeNet, MobileNetV2, and ShuffleNet deep learning models with SGDM Optimizer and SVM classifier yielded the best results for MICC-F220 dataset. Fine-tuned VGG19, MobileNetV2, and ResNet-50 deep learning models with SGDM Optimizer and SVM v classifier yielded the best results for Columbia dataset. In CoMoFoD dataset, fine-tuned AlexNet, MobileNetV2, and ShuffleNet deep learning models with SGDM Optimizer and SVM classifier yielded the best results. The proposed approach, utilizing machine learning algorithms and deep learning features, enhanced forgery detection and reduced false positives. Results were validated on benchmark image forgery datasets and compared to current methods.

Advancing glioma diagnosis: Integrating custom U-Net and VGG-16 for improved grading in MR imaging.

In the realm of medical imaging, the precise segmentation and classification of gliomas represent fundamental challenges with profound clinical implications. Leveraging the BraTS 2018 dataset as a standard benchmark, this study delves into the potential of advanced deep learning models for addressing these challenges. We propose a novel approach that integrates a customized U-Net for segmentation and VGG-16 for classification. The U-Net, with its tailored encoder-decoder pathways, accurately identifies glioma regions, thus improving tumor localization. The fine-tuned VGG-16, featuring a customized output layer, precisely differentiates between low-grade and high-grade gliomas. To ensure consistency in data pre-processing, a standardized methodology involving gamma correction, data augmentation, and normalization is introduced. This novel integration surpasses existing methods, offering significantly improved glioma diagnosis, validated by high segmentation dice scores (WT: 0.96, TC: 0.92, ET: 0.89), and a remarkable overall classification accuracy of 97.89%. The experimental findings underscore the potential of integrating deep learning-based methodologies for tumor segmentation and classification in enhancing glioma diagnosis and formulating subsequent treatment strategies.

Comparative Analysis of VGG-16 and ResNet-50 for Occluded Ear Recognition

Occluded ear recognition is a challenging task in biometric systems due to the presence of occlusions that can hinder accurate identification. There is still a research gap in enhancing the robustness of deep learning to handle severities of occlusions with different datasets. This research focuses on developing a robust occluded ear recognition system by implementing fine-tuning techniques on three popular pre-trained deep learning models, Residual Neural Network (ResNet-50), Visual Geometry Group (VGG-16), and EfficientNet. The system is evaluated on two manually occluded ear datasets, which are the AMI ear dataset and the IITD ear dataset. The experiment results showed the fine-tuned ResNet-50 model performs better than the fine-tuned VGG-16 model. The results indicate that the model's ability to accurately predict the classes or labels decreases as more data is occluded. Higher occlusion rates lead to a loss of important information, making it more challenging for the model to distinguish between different patterns and make accurate predictions. According to the findings, the amount of occlusion influenced the identification accuracy and worsened as the occlusion became larger. In the future, ear recognition systems will likely continue to improve in accuracy and be adopted by a wider range of organizations and industries. They may also be integrated with other biometric technologies and used for personalization purposes. However, ethical considerations related to the use of ear recognition systems will also need to be addressed.

Comparative Analysis of VGG-16 and ResNet-50 for Occluded Ear Recognition

Occluded ear recognition is a challenging task in biometric systems due to the presence of occlusions that can hinder accurate identification. There is still a research gap in enhancing the robustness of deep learning to handle severities of occlusions with different datasets. This research focuses on developing a robust occluded ear recognition system by implementing fine-tuning techniques on three popular pre-trained deep learning models, Residual Neural Network (ResNet-50), Visual Geometry Group (VGG-16), and EfficientNet. The system is evaluated on two manually occluded ear datasets, which are the AMI ear dataset and the IITD ear dataset. The experiment results showed the fine-tuned ResNet-50 model performs better than the fine-tuned VGG-16 model. The results indicate that the model's ability to accurately predict the classes or labels decreases as more data is occluded. Higher occlusion rates lead to a loss of important information, making it more challenging for the model to distinguish between different patterns and make accurate predictions. According to the findings, the amount of occlusion influenced the identification accuracy and worsened as the occlusion became larger. In the future, ear recognition systems will likely continue to improve in accuracy and be adopted by a wider range of organizations and industries. They may also be integrated with other biometric technologies and used for personalization purposes. However, ethical considerations related to the use of ear recognition systems will also need to be addressed.

SDIF-CNN: Stacking deep image features using fine-tuned convolution neural network models for real-world malware detection and classification

The detection of malware is a complex problem in the area of Internet security. Developing a malware defense system that is less costly to detect large-scale malware is needed. This paper proposes a novel malware detection and classification architecture based on image visualization as SDIF-CNN: Stacking deep image features using fine-tuned convolution neural networks. The hybrid methodology of transfer learning as fine-tuning and feature extractor of deep convolution neural network models is designed. At first, the pre-trained VGG16 CNN model is deeply fine-tuned with different hyperparameters, including the number of layers, learning rate, momentum, etc. The transfer learning-based fine-tuned VGG16 model is used as a feature extractor along with the three similar pre-trained CNN models, VGG19, ResNet50, and InceptionV3, to obtain the diverse feature map. The extracted features are horizontally concatenated to construct a single feature map. The different feature selection methodologies, including filter-based methods and embedded methods, such as linear regression and random forest, are designed to discard the irrelevant features from a stacked feature map. After that, this study uses six machine learning and deep learning classifiers- K-Nearest Neighbor (K-NN), Support Vector Machine (SVM), Random Forest (RF), Multi-Layer Perceptron (MLP), Extra Tree (ET), and Gaussian Naive Bayes (GNB) by using the stacked feature map as a training feature vector. The hyperparameter optimization of the MLP model as the best classifier is performed using a randomized search algorithm to devise an optimal classifier. The experiments are performed using a publicly benchmarked MalImg dataset of 9339 images from 25 families. The model is also validated on real-world and packed malicious programs to prove the generalization of the proposed methodology in detecting real-world malware. In the proposed system, the MLP model obtained the best performance results as 98.55% accuracy, 99% precision, 99% recall, and 99% F1-score for MalImg datasets, and accuracy of 94.78% for real-world malware datasets. The proposed methodology is resilient to commonly used obfuscation techniques and does not depend upon code disassembly, reverse engineering analysis, and highly resource-intensive dynamic analysis.

Enhancement of VGG16 model with multi-view and spatial dropout for classification of mosquito vectors.

Mosquitoes transmit pathogens that can cause numerous significant infectious diseases in humans and animals such as malaria, dengue fever, chikungunya fever, and encephalitis. Although the VGG16 model is not one of the most advanced CNN networks, it is reported that a fine-tuned VGG16 model achieves accuracy over 90% when applied to the classification of mosquitoes. The present study sets out to improve the accuracy and robustness of the VGG16 network by incorporating spatial dropout layers to regularize the network and by modifying its structure to incorporate multi-view inputs. Herein, four models are implemented: (A) early-combined, (B) middle-combined, (C) late-combined, and (D) ensemble model. Moreover, a structure for combining Models (A), (B), (C), and (D), known as the classifier, is developed. Two image datasets, including a reference dataset of mosquitoes in South Korea and a newly generated dataset of mosquitoes in Thailand, are used to evaluate our models. Regards the reference dataset, the average accuracy of ten runs improved from 83.26% to 99.77%, while the standard deviation decreased from 2.60% to 0.12%. When tested on the new dataset, the classifier's accuracy was also over 99% with a standard deviation of less than 2%. This indicates that the algorithm achieves high accuracy with low variation and is independent of a particular dataset. To evaluate the robustness of the classifier, it was applied to a small dataset consisting of mosquito images captured under various conditions. Its accuracy dropped to 86.14%, but after retraining with the small dataset, it regained its previous level of precision. This demonstrates that the classifier is resilient to variation in the dataset and can be retrained to adapt to the variation. The classifier and the new mosquito dataset could be utilized to develop an application for efficient and rapid entomological surveillance for the prevention and control of mosquito-borne diseases.

Data Augmentation-based Novel Deep Learning Method for Deepfaked Images Detection

Recent advances in artificial intelligence have led to deepfake images, enabling users to replace a real face with a genuine one. deepfake images have recently been used to malign public figures, politicians, and even average citizens. deepfake but realistic images have been used to stir political dissatisfaction, blackmail, propagate false news, and even carry out bogus terrorist attacks. Thus, identifying real images from fakes has got more challenging. To avoid these issues, this study employs transfer learning and data augmentation technique to classify deepfake images. For experimentation, 190,335 RGB-resolution deepfake and real images and image augmentation methods are used to prepare the dataset. The experiments use the deep learning models: convolutional neural network (CNN), Inception V3, visual geometry group (VGG19) and VGG16 with a transfer learning approach. Essential evaluation metrics (accuracy, precision, recall, F1-score, confusion matrix and AUC-ROC curve score) are used to test the efficacy of the proposed approach. Results revealed that the proposed approach achieves an accuracy, recall, F1-score and AUC-ROC score of 90% and 91% precision, with our fine-tuned VGG16 model outperforming other DL models in recognizing real and deepfakes.

Analysis of facial emotion expression in eating occasions using deep learning

Eating is experienced as an emotional social activity in any culture. There are factors that influence the emotions felt during food consumption. The emotion felt while eating has a significant impact on our lives and affects different health conditions such as obesity. In addition, investigating the emotion during food consumption is considered a multidisciplinary problem ranging from neuroscience to anatomy. In this study, we focus on evaluating the emotional experience of different participants during eating activities and aim to analyze them automatically using deep learning models. We propose a facial expression-based prediction model to eliminate user bias in questionnaire-based assessment systems and to minimize false entries to the system. We measured the neural, behavioral, and physical manifestations of emotions with a mobile app and recognize emotional experiences from facial expressions. In this research, we used three different situations to test whether there could be any factor other than the food that could affect a person’s mood. We asked users to watch videos, listen to music or do nothing while eating. This way we found out that not only food but also external factors play a role in emotional change. We employed three Convolutional Neural Network (CNN) architectures, fine-tuned VGG16, and Deepface to recognize emotional responses during eating. The experimental results demonstrated that the fine-tuned VGG16 provides remarkable results with an overall accuracy of 77.68% for recognizing the four emotions. This system is an alternative to today’s survey-based restaurant and food evaluation systems.

A comparative study of deep learning and Internet of Things for precision agriculture

Precision farming is made possible by rapid advances in deep learning (DL) and the internet of things (IoT) for agriculture, allowing farmers to upgrade their agriculture operations to sustainably fulfill the future food supply. This paper presents a comprehensive overview of recent research contributions in DL and IoT for precision agriculture. This paper surveys the diverse research on DL applications in agriculture, such as detecting pests, disease, yield, weeds, and soil, including fundamental DL techniques. Also, the work describes the IoT architecture and analyzes sensor categorization, agriculture sensors, and unmanned arial vehicles (UAVs) used in recent research. Besides that, data acquisition, annotation, and augmentation for agriculture datasets were covered, and a few widely used datasets were listed. This work also discusses some challenges and issues that DL and IoT face. Furthermore, the research proposed a bootstrapping approach of Transfer learning where fine-tuned VGG16 is fused with optimized and improved newly built fully connected layers for pest detection. The performance of the proposed model is evaluated and compared with other models, such as custom VGG16 as a classifier; fine-tuned VGG16 is optimized with other optimizers like SGD, RMSProp, and Adam. The results show that the proposed model for pest detection outperforms all other models with an accuracy of 96.58 % and a loss of 0.15%. The review and the proposed work presented in this paper will significantly direct researchers toward DL and IoT for intelligent farming.

Deep learning for the diagnosis of mesial temporal lobe epilepsy.

This study aimed to enable the automatic detection of the hippocampus and diagnose mesial temporal lobe epilepsy (MTLE) with the hippocampus as the epileptogenic area using artificial intelligence (AI). We compared the diagnostic accuracies of AI and neurosurgical physicians for MTLE with the hippocampus as the epileptogenic area. In this study, we used an AI program to diagnose MTLE. The image sets were processed using a code written in Python 3.7.4. and analyzed using Open Computer Vision 4.5.1. The deep learning model, which was a fine-tuned VGG16 model, consisted of several layers. The diagnostic accuracies of AI and board-certified neurosurgeons were compared. AI detected the hippocampi automatically and diagnosed MTLE with the hippocampus as the epileptogenic area on both T2-weighted imaging (T2WI) and fluid-attenuated inversion recovery (FLAIR) images. The diagnostic accuracies of AI based on T2WI and FLAIR data were 99% and 89%, respectively, and those of neurosurgeons based on T2WI and FLAIR data were 94% and 95%, respectively. The diagnostic accuracy of AI was statistically higher than that of board-certified neurosurgeons based on T2WI data (p = 0.00129). The deep learning-based AI program is highly accurate and can diagnose MTLE better than some board-certified neurosurgeons. AI can maintain a certain level of output accuracy and can be a reliable assistant to doctors.

A two-phase approach for expression invariant 3D face recognition using fine-tuned VGG-16 and 3D-SIFT descriptors

A two-phase approach for expression invariant 3D face recognition using fine-tuned VGG-16 and 3D-SIFT descriptors

Diagnosis of COVID-19 from Multimodal Imaging Data Using Optimized Deep Learning Techniques.

COVID-19 had a global impact, claiming many lives and disrupting healthcare systems even in many developed countries. Various mutations of the severe acute respiratory syndrome coronavirus-2, continue to be an impediment to early detection of this disease, which is vital for social well-being. Deep learning paradigm has been widely applied to investigate multimodal medical image data such as chest X-rays and CT scan images to aid in early detection and decision making about disease containment and treatment. Any method for reliable and accurate screening of COVID-19 infection would be beneficial for rapid detection as well as reducing direct virus exposure in healthcare professionals. Convolutional neural networks (CNN) have previously proven to be quite successful in the classification of medical images. A CNN is used in this study to suggest a deep learning classification method for detecting COVID-19 from chest X-ray images and CT scans. Samples from the Kaggle repository were collected to analyse model performance. Deep learning-based CNN models such as VGG-19, ResNet-50, Inception v3 and Xception models are optimized and compared by evaluating their accuracy after pre-processing the data. Because X-ray is a less expensive process than CT scan, chest X-ray images are considered to have a significant impact on COVID-19 screening. According to this work, chest X-rays outperform CT scans in terms of detection accuracy. The fine-tuned VGG-19 model detected COVID-19 with high accuracy-up to 94.17% for chest X-rays and 93% for CT scans. This work thereby concludes that VGG-19 was found to be the best suited model to detect COVID-19 and chest X-rays yield better accuracy than CT scans for the model.

Breast Cancer Detection Using Breastnet-18 Augmentation with Fine Tuned Vgg-16

Women from middle age to old age are mostly screened positive for Breast cancer which leads to death. Times over the past decades, the overall survival rate in breast cancer has improved due to advancements in early-stage diagnosis and tailored therapy. Today all hospital brings high awareness and early detection technologies for breast cancer. This increases the survival rate of women. Though traditional breast cancer treatment takes so long, early cancer techniques require an automation system. This research provides a new methodology for classifying breast cancer using ultrasound pictures that use deep learning and the combination of the best characteristics. Initially, after successful learning of Convolutional Neural Network (CNN) algorithms, data augmentation is used to enhance the representation of the feature dataset. Then it uses BreastNet18 with fine-tuned VGG-16 model for pre-training the augmented dataset. For feature classification, Entropy controlled Whale Optimization Algorithm (EWOA) is used. The features that have been optimized using the EWOA were utilized to fuse and optimize the data. To identify the breast cancer pictures, training classifiers are used. By using the novel probability-based serial technique, the best-chosen characteristics are fused and categorized by machine learning techniques. The main objective behind the research is to increase tumor prediction accuracy for saving human life. The testing was performed using a dataset of enhanced Breast Ultrasound Images (BUSI). The proposed method improves the accuracy compared with the existing methods.

Nucleus segmentation and classification using residual SE-UNet and feature concatenation approach incervical cytopathology cell images.

Introduction: Pap smear is considered to be the primary examination for the diagnosis of cervical cancer. But the analysis of pap smear slides is a time-consuming task and tedious as it requires manual intervention. The diagnostic efficiency depends on the medical expertise of the pathologist, and human error often hinders the diagnosis. Automated segmentation and classification of cervical nuclei will help diagnose cervical cancer in earlier stages. Materials and Methods: The proposed methodology includes three models: a Residual-Squeeze-and-Excitation-module based segmentation model, a fusion-based feature extraction model, and a Multi-layer Perceptron classification model. In the fusion-based feature extraction model, three sets of deep features are extracted from these segmented nuclei using the pre-trained and fine-tuned VGG19, VGG-F, and CaffeNet models, and two hand-crafted descriptors, Bag-of-Features and Linear-Binary-Patterns, are extracted for each image. For this work, Herlev, SIPaKMeD, and ISBI2014 datasets are used for evaluation. The Herlev datasetis used for evaluating both segmentation and classification models. Whereas the SIPaKMeD and ISBI2014 are used for evaluating the classification model, and the segmentation model respectively. Results: The segmentation network enhanced the precision and ZSI by 2.04%, and 2.00% on the Herlev dataset, and the precision and recall by 0.68%, and 2.59% on the ISBI2014 dataset. The classification approach enhanced the accuracy, recall, and specificity by 0.59%, 0.47%, and 1.15% on the Herlev dataset, and by 0.02%, 0.15%, and 0.22% on the SIPaKMed dataset. Conclusion: The experiments demonstrate that the proposed work achieves promising performance on segmentation and classification in cervical cytopathology cell images..

Non-destructive Assessment of White Striping in Broiler Breast Meat Using Structured-Illumination Reflectance Imaging with Deep Learning

Highlights Broiler breast meat with white striping (WS) was imaged under sinusoidally modulated structured illumination. Amplitude component (AC) images resolved better WS characteristics than direct component (DC) images. Models built on deep features classified meat samples into 2 and 3 classes according to WS severity. The AC images consistently outperformed the DC images and achieved the best accuracy of 96.6% in 2-class modeling. Abstract. Visual inspection is the prevailing practice in the industry for assessing white striping (WS) in broiler breast meat. However, this approach is subjective, laborious, and prone to error. Several studies have utilized imaging technology under uniform illumination; however, detecting defects, such as WS, remains challenging. This study investigated the efficacy of the emerging structured illumination reflectance imaging (SIRI) combined with deep learning (DL) for the WS assessment of broiler meat. Broiler fillets with varying degrees of WS were imaged using a custom-assembled monochromatic SIRI system (0.05-0.40 cycles/mm). The acquired SIRI pattern images were demodulated into direct components (DC) and amplitude components (AC) at each spatial frequency. Pre-trained DL models, including two Very Deep Convolutional Networks (VGG16 and VGG19) and two Densely Connected Convolutional Neural Networks (DenseNet121 and DensetNet201), were evaluated as fine-tuned end-to-end classifiers and feature extractors separately to differentiate the meat samples. Fine-tuned VGG16 resulted in the best 2-class and 3-class classification accuracies of 94.5% and 74.6%, respectively, based on the AC images at 0.30 cycles/mm, improving over the accuracies of the corresponding DC images by 10.4% and 8.9%, respectively. Fine-tuned VGG19 and DenseNet201 also achieved substantial improvements of AC over DC images by 12% or higher. Linear discriminant analysis, in conjunction with principal component analysis for dimension reduction, yielded better accuracies of 96.6% in 2-class and 83.4% in 3-class classification, based on the deep features extracted by DenseNet121 from the AC images at 0.40 cycles/mm and 0.30 cycles/mm, respectively, representing improvements of 5.0% and 2.9% based on the features of the corresponding DC images. The SIRI technique combined with DL is effective for differentiating between normal and WS-affected broiler meat. Keywords: Deep learning, Imaging, Meat quality, Poultry, Structured illumination, White striping.

Malaria Blood Cell Image Classification using Transfer Learning with Fine-Tune ResNet50 and Data Augmentation

Based on the WHO Report related to malaria, it is estimated that there will be 241 million malaria cases and 627,000 deaths from this disease globally in 2020 with the number of deaths increasing yearly. Preventing malaria disease conditions is through early detection. A more quick and precise malaria diagnosis method was required to simplify and reduce the detection process. Medical image classification could be carried out rapidly and precisely using machine learning or deep learning techniques. This research aims to diagnose malaria by classifying images of malaria blood cells using Deep Learning with a Transfer Learning approach. By utilizing various fine-tuning procedures and implementing data augmentation proposed method develops the method from previous studies. Two types of models Frozen ResNet50 and Fine-Tune ResNet50 are being tested. The dataset utilized will be augmented to improve model performance. This study makes use of the "NIH Malaria Cell Images Dataset" a dataset that contains a total of 27,660 image data. It is divided into two classes: parasitized and uninfected. The results are improved from previous research using the fine-tuned VGG16 model with an accuracy of 96% compared to this study using the fine-tuned ResNet50 model which achieved an accuracy score of 98%.

Fault diagnosis of photovoltaic modules using deep neural networks and infrared images under Algerian climatic conditions

The number of decentralized photovoltaic (PV) systems generating electricity has increased significantly, and its monitoring and maintenance has become a challenge in terms of stability, reliability, security, efficiency, as well as energy production costs. Hence, prevention against faults and breakdowns becomes essential. In this work, a Convolutional Neural Network (CNN) model and a fine-tuned model based on Visual Geometry Group (VGG-16) have been examined to address the issue of fault diagnosis of PV modules using thermographic images. For fault detection, we have used binary classification, and multiclass classification for identification the type of fault. The database used in this study was made up of an imbalanced class distribution of infrared thermographic images of PV modules under normal and faulty conditions (such as bypass diode failure, partially covered PV module, shading effect, short-circuit and dust deposit on the PV surface). The test facility is located at the Unit for Developing Solar Equipment's (UDES), in the north of Algeria. The average accuracy archived using the fine-tuned VGG-16 model is 99.91% for the fault detection and 99.80% for the fault diagnosis of five types of defects. Experimental tests show high accurate prediction results using the fine-tuned model and somewhat less accuracy using the small Deep Convolutional Neural Network (small-DCNN) model.