Fig 8 - uploaded by Clifford K. Yang
Content may be subject to copyright.
Illustration of the network architecture of VGG-19 model: conv means convolution, FC means fully connected
Context in source publication
Context 1
... and Zisserman of the University of Oxford created a 19-layer (16 conv., 3 fully-connected) CNN that strictly used 3×3 filters with stride and pad of 1, along with 2×2 max-pooling layers with stride 2, called VGG-19 model. 28,29 Compared to AlexNet, the VGG-19 (see Fig. 8) is a deeper CNN with more layers. To reduce the number of parameters in such deep networks, it uses small 3×3 filters in all convolutional layers and best utilized with its 7.3% error rate. The VGG-19 model was not the winner of ILSVRC 30 2014, however, the VGG Net is one of the most influential papers because it reinforced the notion ...
Similar publications
Background
Mammography is the modality of choice for breast cancer screening. However, some cases of breast cancer have been diagnosed through ultrasonography alone with no or benign findings on mammography (hereby referred to as non-visibles). Therefore, this study aimed to identify factors that indicate the possibility of non-visibles based on th...
Purpose
Despite the controversy surrounding the role of clinical breast exam (CBE) in modern breast cancer screening, it is widely practiced. We examined the contribution of CBE in women undergoing routine screening mammography and in women under the screening age.
Methods
A retrospective cohort study including all women participating in a volunta...
Purpose
Breast cancer screening programs often encounter challenges related to high costs and limited accessibility, particularly in rural areas. Mobile mammography units (MMUs) have emerged as a potential solution to address these barriers. This study aimed to compare the outcomes of mammography screenings conducted in fixed units and MMUs, focusi...
While two-view mammography taking both mediolateral-oblique (MLO) and cranio-caudual (CC) views is the current standard method of examination in breast cancer screening, single-view mammography is still being performed in some countries on women of specific ages. The rate of cancer detection is lower with single-view mammography than for two-view m...
Artificial intelligence (AI) improves the accuracy of mammography screening, but prospective evidence, particularly in a single-read setting, remains limited. This study compares the diagnostic accuracy of breast radiologists with and without AI-based computer-aided detection (AI-CAD) for screening mammograms in a real-world, single-read setting. A...
Citations
... Its hierarchical structure allows the modeling of patterns from simple edges to complex structures, which is important in applications such as keratoconus detection in medical imaging. The architecture of VGG19 is shown in Figure 12 [59]. ...
Keratoconus is a progressive eye disorder that, if undetected, can lead to severe visual impairment or blindness, necessitating early and accurate diagnosis. The primary objective of this research is to develop a feature fusion hybrid deep learning framework that integrates pretrained Convolutional Neural Networks (CNNs) with Vision Transformers (ViTs) for the automated classification of keratoconus into three distinct categories: Keratoconus, Normal, and Suspect. The dataset employed in this study is sourced from a widely recognized and publicly available online repository. Prior to model development, comprehensive preprocessing techniques were applied, including the removal of low-quality samples, image resizing, rescaling, and data augmentation. The dataset was subsequently partitioned into training, testing, and validation subsets to facilitate robust model training and performance evaluation. Eight state-of-the-art CNN architectures, including DenseNet121, EfficientNetB0, InceptionResNetV2, InceptionV3, MobileNetV2, ResNet50, VGG16, and VGG19, were utilized for feature extraction, while the ViT served as the classification head, leveraging its global attention mechanism for enhanced contextual learning, achieving near-perfect accuracy (e.g., DenseNet121+ViT: 99.28%). This study underscores the potential of hybrid CNN-ViT architectures to revolutionize keratoconus diagnosis, offering scalable, accurate, and efficient solutions to overcome limitations of traditional diagnostic methods while paving the way for broader applications in medical imaging. Doi: 10.28991/ESJ-2025-09-02-027 Full Text: PDF
... To decrease the number of parameters in this CNN network, small 3 × 3 filters are used in all convolutional layers. In the Imagenet competition in 2014, VGG19 was trained on more than one million images and learned rich feature representations for a large range of images [72]. ...
Detecting camouflaged objects in camouflage images is quite challenging due to their closely matching texture, pattern, and color characteristics with the background. Existing binary segmentation solutions cannot easily deal with the problem of detecting camouflaged objects because they have weak boundaries and background-like patterns. The purpose of camouflaged object detection (COD) is to detect objects that very closely resemble the background. In this study, an original camouflage butterfly dataset called ERVA 1.0 is created, consisting of images of 10 butterfly species downloaded from search engines. Additionally, the raw training data is increased with data augmentation techniques. For COD, this study presents a two-stage solution: segmentation and object recognition. The texture features of all test images on the ERVA 1.0 dataset are extracted utilizing the Gabor filter for segmentation. Then, these extracted features are clustered with the K-means algorithm, and the original image is separated into different regions based on texture features. The local binary pattern algorithm and Euclidean distance calculation are used to determine which of these regions belongs to the butterfly object. Following the application of morphological operations on the identified butterfly object region, pretrained models from deep learning techniques were employed to predict the species of the butterfly. Segmentation success rates are 87.89% with the structural similarity method and 83.64% with the Dice similarity coefficient method. Deep learning pretrained models are used to classify the type of the butterfly object obtained after segmentation. Experiment 1 was conducted with un-augmented training data and Experiment 2 with augmented data by applying data augmentation techniques. The highest success rate for Experiment 1 was 92.29% with the InceptionResNetV2 model, and the highest success rate for Experiment 2 is 94.81% with the DenseNet121 model.
... The architecture network of VGG-19 model[17]. ...
... The architecture network of VGG-19 model[17]. ...
In the aftermath of catastrophic natural catastrophes such as earthquakes, tsunamis, and explosions, providing immediate help to key areas can mean the difference between life and death for many individuals. To meet this critical demand, we created a hybrid transportation system that harnesses the power of vgg19 and traditional shortest path algorithms. The objective was to create a real-time system that could address these issues and provide a novel viewpoint. The suggested approach can precisely anticipate damaged roads and steer clear of them when determining the shortest way between sites during emergencies or natural disasters by merging VGG19 with the shortest path algorithm. With the potential to save many lives, this creative strategy can assist emergency responders reach vital places swiftly and effectively while also saving important time and resources. The experimental study shows that the proposed model can achieve robust results. In fact, our solution achieves 98% accuracy rate and a 0.972 G-Mean score on the test set.
... This model has a total of 138 million parameters. VGG19 is pre-trained on a subset of the ImageNet dataset [13], which is used in the ImageNet Large-Scale Visual Recognition Challenge, containing over a million images, enabling it to classify images across thousands of classes [13]. This extensive training has equipped VGG19 with the ability to capture detailed and varied feature representations from a wide range of images, thereby making it highly effective in feature extraction tasks. ...
... This model has a total of 138 million parameters. VGG19 is pre-trained on a subset of the ImageNet dataset [13], which is used in the ImageNet Large-Scale Visual Recognition Challenge, containing over a million images, enabling it to classify images across thousands of classes [13]. This extensive training has equipped VGG19 with the ability to capture detailed and varied feature representations from a wide range of images, thereby making it highly effective in feature extraction tasks. ...
In the fashion retail e-commerce sector, personalized product recommendations are crucial for enhancing the shopping experience. This study introduces a method that combines a pre-trained deep learning model named VGG19 with the 10 nearest neighbors algorithm to recommend visually similar products. VGG19 is utilized to extract detailed features from product images, enabling more accurate recommendations. The nearest neighbors algorithm then selects the ten products most similar to those previously viewed by customers. Recommendations are ranked based on customer purchase frequency to prioritize the most popular and relevant items. This method's practical applicability was demonstrated by testing it on a diverse set of products, including jackets from outerwear, baby bodysuits from children's wear, socks from footwear, and sunglasses from the accessories category.
... The architecture network of VGG-19 model[17]. ...
In the aftermath of catastrophic natural catastrophes such as earthquakes, tsunamis, and explosions, providing immediate help to key areas can mean the difference between life and death for many individuals. To meet this critical demand, we created a hybrid transportation system that harnesses the power of vgg19 and traditional shortest path algorithms. The objective was to create a real-time system that could address these issues and provide a novel viewpoint. The suggested approach can precisely anticipate damaged roads and steer clear of them when determining the shortest way between sites during emergencies or natural disasters by merging VGG19 with the shortest path algorithm. With the potential to save many lives, this creative strategy can assist emergency responders reach vital places swiftly and effectively while also saving important time and resources. The experimental study shows that the proposed model can achieve robust results. In fact, our solution achieves 98% accuracy rate and a 0.972 G-Mean score on the test set.
... The VGG-19 network contains more parameters (138 million) compared to the VGG-16 network in parallel with its approximate number of layers [21]. ...
Using lung images obtained by computed tomography (CT), this study aims to detect coronavirus (Covid-19) disease with deep learning (DL) techniques. The study included 751 lung CT images from 118 Covid-19 patients and 628 lung CT images from 100 healthy individuals. In total, 70% of the 1379 images were used for training and 30% for testing. In the study, two different methods were proposed on the same dataset. In the first method, the images were trained on AlexNet, VGG-16, VGG-19, GoogleNet and a proposed network. The performance metrics obtained from the five networks were compared and it was observed that the proposed network achieved the highest accuracy value with 95.61%. In the second method, the images were trained on VGG-16, VGG-19, DenseNet-121, ResNet-50 and MobileNet networks. Among the image features obtained from each of these networks, the best 1000 features were selected by Principal Component Analysis (PCA). The best 1000 features were classified with Random Forest (RF) and Support Vector Machines (SVM). According to the classification results, the best 1000 features selected from the features extracted by the VGG-16 and MobileNet networks were obtained with the highest accuracy rate of 93.94% using SVM. It is thought that this study can be a helpful tool in the diagnosis of Covid-19 disease while reducing time and labor costs with the use of artificial intelligence (AI).
... Deep Convolutional Neural Networks (DCNNs), deep learning technique have the ability to automatically learn features through several network layers from large labelled datasets. In medical image analysis, DCNNs have been successfully utilized for various tasks such as skin lesion classification [2], suspicious region detection [3] classification of diabetic retinopathy on fundoscopic images [4], and automatic classification of breast cancer histopathological images [5]. ...
The recent release of large amounts of Chest radiographs (CXR) has prompted the research of automated analysis of Chest X-rays to improve health care services. DCNNs are well suited for image classification because they can learn to extract features from images that are relevant to the task at hand. However, class imbalance is a common problem in chest X-ray imaging, where the number of samples for some disease category is much lower than the number of samples in other categories. This can occur as a result of rarity of some diseases being studied or the fact that only a subset of patients with a particular disease may undergo imaging. Class imbalance can make it difficult for Deep Convolutional Neural networks (DCNNs) to learn and make accurate predictions on the minority classes. Obtaining more data for minority groups is not feasible in medical research. Therefore, there is a need for a suitable method that can address class imbalance. To address class imbalance in DCNNs, this study proposes, Deep Convolutional Neural Networks with Augmentation. The results show that data augmentation can be applied to imbalanced dataset to increase the representation of the minority class by generating new images that are a slight variation of the original CXR images. This study further evaluates identifiability and consistency of the proposed model.
... The output is generated using the softmax activation function and 1000 fully connected layers [13]. In Fig. 4 [14], the overall architecture of Vgg19 is displayed. ...
Since many years ago, walnuts have been extensively available around the world and come in various quality varieties. The proper variety of walnut can be grown in the right area and is vital to human health. This fruit's production is time-consuming and expensive. However, even specialists find it challenging to differentiate distinct kinds since walnut leaves are so similar in color and feel. There aren't many studies on the classification of walnut leaves in the literature, and the most of them were conducted in laboratories. The classification process can now be carried out automatically from leaf photos thanks to technological advancements. The walnut data set was applied to the suggested deep learning model. There aren't many studies on the classification of walnut leaves in the literature, and the most of them were conducted in laboratories. The walnut data set, which consists of 18 different types of 1751 photos, was used to test the suggested deep learning model. The three most successful algorithms among the commonly utilized CNN algorithms in the literature were first selected for the suggested model. From the Vgg16, Vgg19, and AlexNet CNN algorithms, many features were retrieved. Utilizing the Whale Optimization Algorithm (WOA), a new feature set was produced by choosing the top extracted features. KNN is used to categorize this feature set. An accuracy rating of 92.59% was attained as a consequence of the tests.
... To verify the superiority of the proposed method in terms of feature extraction and diagnostic performance, we compared it with traditional machine learning methods and state-of-theart deep learning methods. The compared methods include shallow CNN, VGG19 [37], GoogLeNet [38] and DenseNet [39]. To realize zero-shot compound mechanical fault diagnosis using these methods, we retrained these models respectively from scratch using different data input methods of CB vibration signals shown in table 1, and constructed fault ALs similar to our proposed method. ...
Diagnosis of compound mechanical faults for power circuit breakers (CBs) is a challenging task. In traditional fault diagnosis methods, however, all fault types need to be collected in advance for the training of diagnosis model. Such processes have poor generalization capabilities for industrial scenarios with no or few data when faced with new faults. In this study, we propose a novel zero-shot learning method named DSR-AL to address this problem. An unsupervised neural network, namely, depthwise separable residual convolutional neural network (DSRCNN), is designed to directly learn features from 3D time-frequency images of CB vibration signals. Then we build fault attribute learners (ALs), for transferring fault knowledge to the target faults. Finally, the ALs are used to predict the attribute vector of the target faults, thus realizing the recognition of previously unseen faults. The orthogonal experiments are designed and conducted on real industrial switchgear to validate the effectiveness of the proposed diagnosis framework. Results show that it is feasible to diagnose target faults without using their samples for training, which greatly saves the costs of collecting fault samples. This will help to accurately identify the various faults that may occur during CB's life cycle, and facilitate the application of intelligent fault diagnosis system.
