VGG16 Model Research Articles

Abstract B010: Improving pancreatic cancer image classification with transfer learning: CD45 vs. PANCK

Abstract Pancreatic ductal adenocarcinoma (PDAC) is known for its high aggressiveness and is the fourth most common cause of death of cancer-related death in the United States. It is highly malignant and often goes undiagnosed until advanced stages due to its late-onset symptoms, resulting in poor prognosis. Using quantitative analyses of histological features, such as the expression of prognostic biomarker, can provide valuable insight to guide treatment decisions, monitor disease progression and predict outcomes. Previous researchers have found that CD45 and PANCK play a huge role in characterizing the tumor microenvironment of PDAC. One study using single-cell RNA sequencing compared the microenvironments of healthy pancreata and pancreatic tumors, revealing differences in cellular composition. Tumor samples showed a higher proportion of immune cells, including CD45+ cells, compared to heathy samples and utilized PANCK to differentiate between epithelial and non-epithelial cells. This study aims to enhance the accuracy of classifying pancreatic cancer pathological images based on prognostic markers using transfer learning techniques. We curated and pre-processed a dataset of pancreatic cancer pathological images with different prognostic markers. Using the pre-trained convolutional neural network (CNN) model VGG16, we fine-tuned it with our dataset to better differentiate between two prognostic markers, CD45 and PANCK. The model's performance was evaluated using a similarity matrix to illustrate distinctions among the two prognostic markers. Our results demonstrate that the transfer learning approach significantly improves classification accuracy, achieving confusion matrix validation results of 90% between CD45 and PANCK prognostic markers. SSIM values were relatively low, highlighting the ability of our model to differentiate between the two prognostic markers. The low SSIM value also shows variability within the two prognostic markers, indicating heterogeneity in structural changes due to pancreatic cancer. These findings indicate that transfer learning can be a powerful tool for improving the classification of pathological images and that there is notable variability in the structural features associated with different prognostic markers in pancreatic cancer. Citation Format: Miracle Thomas. Improving pancreatic cancer image classification with transfer learning: CD45 vs. PANCK [abstract]. In: Proceedings of the AACR Special Conference in Cancer Research: Tumor-body Interactions: The Roles of Micro- and Macroenvironment in Cancer; 2024 Nov 17-20; Boston, MA. Philadelphia (PA): AACR; Cancer Res 2024;84(22_Suppl):Abstract nr B010.

Multimodal Biometric Authentication System for Military Weapon Access: Face and ECG Authentication

Unimodal or Single factor biometric systems refer to biometric systems that employ only one form of biometric data to authenticate an individual’s identity. These kinds of biometrics are susceptible to higher error rates and security vulnerabilities because it relays on a single trait for authentication. To overcome this, multimodal biometrics method is proposed. Multi-modal biometric system can authenticate more than once and some advantages include; highaccuracy, low error rate, and large population coverage. These biometrics systems increase integrity and privacy since it will contain several biometric features of every customer. So, here designed a multimodal biometrics project utilizing deep learning to enhance authentication security by combining face and Electrocardiogram(ECG) signals. VGG-16 model, a deep learning architecture used to capture complex patterns in accurate individual identification with both ECG and Facial data. The high-resolution convolutional filters capture the intricate details of the face and ECG waveform, ensuring high accuracy in distinguishing different individuals.

Technical Note: Neural Network Architectures for Self-Supervised Body Part Regression Models with Automated Localized Segmentation Application.

The advancement of medical image deep learning necessitates tools that can accurately identify body regions from whole-body scans to serve as an essential pre-processing step for downstream tasks. Typically, these deep learning models rely on labeled data and supervised learning, which is labor-intensive. However, the emergence of self-supervised learning is revolutionizing the field by eliminating the need for labels. The purpose of this study was to compare neural network architectures of self-supervised models that produced a body part regression (BPR) slice score to aid in the development of anatomically localized segmentation models. VGG, ResNet, DenseNet, ConvNext, and EfficientNet BPR models were implemented in the MONAI/Pytorch framework. Landmark organs were correlated to slice scores and mean absolute error (MAE) was calculated from the predicted slice and the actual slice of various organ landmarks. Four localized DynUNet segmentation models (thorax, upper abdomen, lower abdomen, and pelvis) were developed using the BPR slice scores. Dice similarity coefficient (DSC) was compared between the localized and baseline segmentation models. The best performing BPR model was the EfficientNet architecture with an overall 3.18 MAE, compared to the VGG baseline model with a MAE of 6.29. The localized segmentation model significantly outperformed the baseline in 16 out of 20 organs with a DSC of 0.88. Enhanced neural networks like EfficientNet have a large performance increase in localizing anatomical structures in a CT compared in BPR task. Utilizing BPR slice score is shown to be effective in anatomically localized segmentation tasks with improved performance.

Enhancing multiclass COVID-19 prediction with ESN-MDFS: Extreme smart network using mean dropout feature selection technique.

Deep learning and artificial intelligence offer promising tools for improving the accuracy and efficiency of diagnosing various lung conditions using portable chest x-rays (CXRs). This study explores this potential by leveraging a large dataset containing over 6,000 CXR images from publicly available sources. These images encompass COVID-19 cases, normal cases, and patients with viral or bacterial pneumonia. The research proposes a novel approach called "Enhancing COVID Prediction with ESN-MDFS" that utilizes a combination of an Extreme Smart Network (ESN) and a Mean Dropout Feature Selection Technique (MDFS). This study aimed to enhance multi-class lung condition detection in portable chest X-rays by combining static texture features with dynamic deep learning features extracted from a pre-trained VGG-16 model. To optimize performance, preprocessing, data imbalance, and hyperparameter tuning were meticulously addressed. The proposed ESN-MDFS model achieved a peak accuracy of 96.18% with an AUC of 1.00 in a six-fold cross-validation. Our findings demonstrate the model's superior ability to differentiate between COVID-19, bacterial pneumonia, viral pneumonia, and normal conditions, promising significant advancements in diagnostic accuracy and efficiency.

Elevating Facial Expression Detection: Empowered by VGG-19 and Weight-Normalized Gradient Boost Algorithm

Facial expression recognition termed as FER, is one of the vital tasks which is able in mimicking the human ability and wellness. Face expression and other gestures delivered via face are some of the important aspects of conveying non-verbal communications which plays a vital role in interpersonal relations. This domain has a higher scope of research due to enhancing computer vision and human-computer interaction. With respect and consideration to these domain interest, DL has gained a reasonable approach in performing FER. Some of the state-of-art approaches encompassing the ML approaches have faced several cases of laybacks such as overfitting, computational complexity, less adaptable to higher and big datasets. Thus, considering these laybacks and in achieving accurate FER in aspects of bringing proximal levels of FER, the current approach deployed DL methods for both feature extraction from the input image and in classifying them. Deep multi-level feature extraction is performed using the VGG-19 model and Weight Normalized gradient boosting algorithm is adapted for classifying these expressions using the FER13 dataset. This dataset constitutes the input images, which are in ranges of 28,709 sample image data and about 3,589 image data for test. These input images are initially pre-processed for obtaining better accuracy rates when performing feature extraction and classifications. The complete model in effective FER is evaluated using the performance metrics comprising Accuracy (98%), Recall (99%), F1-score (98%) and the precision rates (97%). This analysis of the performance will aid in affirming the overall efficacy of the proposed system.

Defect classification of composite materials using transfer learning methods

ABSTRACT Nowadays, composite materials have become prevalent across various sectors, particularly finding usage in large-scale applications such as spaceships, automobiles, and aircrafts. The accurate detection of the defects in these materials is crucial, yet traditional methods often rely on human inspection, which is susceptible to errors. Recent advancements in machine learning have enabled defect detection using ultrasonic non-destructive testing methods. This paper introduces a new dataset named UNDT, which is obtained from the scans of 60 different composite materials, generating a total of 1150 images depicting both defective and non-defective areas. Several transfer learning methods are applied on the newly introduced UNDT dataset as well as the publicly available USimgAIST ultrasonic dataset. Comparative performance assessments illustrate the significance of utilising the transfer learning approach for defect classification on ultrasonic inspection images. Furthermore, the research emphasises the substantial benefits of employing these transfer learning methods. Notably, the DenseNet121 and VGG19 models achieve the highest accuracy rates, with 98.8% and 98.6% on the UNDT and USimgAIST datasets, respectively.

State identification of Parkinson's disease based on transfer learning.

The local field potential (LFP) signals are a vital signal for studying the mechanisms of deep brain stimulation (DBS) and constructing adaptive DBS containing information related to the motor symptoms of Parkinson's disease (PD). A Parkinson's disease state identification algorithm based on the feature extraction strategy of transfer learning was proposed. The algorithm uses continuous wavelet transform (CWT) to convert one-dimensional LFP signals into two-dimensional gray-scalogram images and color images respectively, and designs a Bayesian optimized random forest (RF) classifier to replace the three fully connected layers for the classification task in the VGG16 model, to realize automatic identification of the pathological state of PD patients. It was found that consistently superior performance of gray-scalogram images over color images. The proposed algorithm achieved an accuracy of 97.76%, precision of 99.01%, recall of 96.47%, and F1-score of 97.73%, outperforming feature extractors such as VGG19, InceptionV3, ResNet50, and the lightweight network MobileNet. This algorithm has high accuracy and can distinguish the disease states of PD patients without manual feature extraction, effectively assisting the working of doctors.

Coupling design features of material surface treatment for ceramic products based on ResNet

Abstract Ceramic products is one of the important carriers of various civilizations, reflecting the lifestyle, aesthetic concepts, and technological level of society at that time. In order to study the surface treatment design features of ceramic craft products, this article analyzed the ceramic features through computer vision technology and used residual neural networks to detect the surface treatment features of ceramic craft products. The extracted texture features were classified to study and analyze the coupling features of different glazes, colors, and shapes on the formation of different textures. This study used ResNeXt50-SSD, which combined ResNeXt50 and SSD (Single Shot MultiBox Detector) algorithms, to compare feature detection with LeNet-5, VGG-16, and MobileNetV2 network models. From the experimental findings, it can be concluded that ResNeXt50-SSD was the most effective for feature recognition of ceramic craft products, with precision, recall, and mAP of 94.3, 92.1, and 89.5%, respectively. Therefore, the combination of ResNeXt50 and SSD algorithms is an effective method for detecting surface treatment features of ceramic craft products.

Potato disease identification using transfer learning approaches

Potato crop is one of the prominent consumed foods by human beings. When potato crops are infected by diseases it affects farmers negatively and to run in a loss. Therefore, early detection of the potato crop disease can play a vital role in minimizing the loss of the farmers. Nowadays, artificial intelligence technologies, more specifically deep learning techniques, provide solutions to many crops disease-related problems. However, training deep learning models requires a high computational power and huge amount of data as they are data hungry models. Also, designing a custom CNN models a difficult task and there are some variations to be considered. To avoid these difficulties, we adopted two pretrained models of DenseNet121 and VGG19 through transfer learning approaches. The achieved accuracy for DenseNet121 and VGG19 models are 82.6% and 98.56% respectively. DenseNet121 model obtained the average precision, sensitivity, and F1-score of 88.19%, 82.53, and 82.04%, respectively. Whereas VGG19 yields 98.39% of precision, 98.39% of sensitivity, and 97.26% of F1-score in ternary 3-class classification (early blight vs healthy vs late blight).

Optimized Deep Learning Model for Medical Image Diagnosis

Deep learning models, particularly convolutional neural networks (CNNs), have excelled in pattern recognition tasks like face recognition, pedestrian detection, and medical diagnosis. CNNs, known for their end-to-end feature extraction and classification, are widely used in computer vision. Pre-trained CNN models such as AlexNet, VGG-19, ResNet, and Inception can serve as feature extractors for various problems, though they often produce a large number of features. To address this challenge and reduce the number of features, this research employs an efficient two-layer feature selection algorithm. The first layer, a global search optimizer, guides the swarm in the search space to identify the optimal feature subset. The second layer refines and fine-tunes the solution obtained from the global search. The main contributions of this work encompass several aspects: (i) the introduction of a lightweight classification model that utilizes a subset of features, (ii) the use of an efficient optimizer to perform simultaneous feature and instance selection, and (iii) the direct implications for the radiology field, where it could serve as a supportive tool for diagnosis. To evaluate the proposed approach, it was tested on skin lesion diagnosis using 1113 abnormal and 1099 normal lesions. The pre-trained VGG-19 model provided a 1-D feature vector of size 4096, with a linear SVM used for classification. Results showed the optimizer reduced the feature vector by up to 86% while achieving over 70% accuracy. The proposed optimizer outperformed the Reinforcement Learning-based Memetic Particle Swarm Optimization (RLMPSO).

Unsupervised Histopathological Sub-Image Analysis for Breast Cancer Diagnosis Using Variational Autoencoders, Clustering, and Supervised Learning

This paper presents an integrated approach to breast cancer diagnosis that combines unsupervised and supervised learning techniques. The method involves using a pre-trained VGG19 model to process sub-images from the BreaKHis dataset, divided into nine parts for comprehensive analysis. This will be followed by a complete description of the architecture and workings of the variational Autoencoder (VAE) used for unsupervised Learning. The encoder network maps the input features to lower dimensions, capturing the most essential information. VAE learns a compressed representation of sub-images, facilitating a more profound understanding of underlying patterns and structures. For this reason, we then employ k-means clustering on the encoded representation to find naturally occurring clusters in our data set comprising a histopathological image. Every single sub-image is later fed into the VGG19-SVM model for classification purposes. During magnification at 100x, this model has attained a fantastic accuracy rate of 98.56%. Combining unsupervised analysis with VAE/k-means clustering and supervised classification with VGG19/SVM can integrate information from both methods, thereby improving the accuracy and robustness of such a task as sub-image classification in breast cancer histopathology.

Harnessing Machine Learning for Effective Waste Classification and Recycling

The world produces nearly 3.5 million tons waste every day so it is important to improve waste management and recycling system. This paper defines how machine learning can classify waste, specifically using VGG16 CNN model to classify 9 types of recyclable materials such as light bulbs, paper, plastic, organic, glass, batteries, clothes, metal and e-waste .We had used nearly 8000+ images from Google Images and Dreamstime.com to test VGG16 model and compare it with another model named as Inceptionv3. But Inceptionv3 showed higher accuracy (~90%) ,and it did not work well when tested on real data. In contrast, the VGG16 model which initially had lower accuracy of (~70%), performed better when tested on real-world scenarios. We also developed a web application which not only classifies the waste but also provides the valuable information to help people how to minimize waste reduction and maximize recycling. The classified images of the waste and its respective findings will also be shared with local authorities to enhance better recycling efforts. Hence this study focuses on need to select right models for waste classification and suggests looking into another machine learning techniques in future.

Deep learning based classification for alzheimer's disease detection using MRI images

Alzheimer's disease is a common type of dementia that can cause serious problems in cognitive functions and activities of daily living. Although there is no definitive cure for Alzheimer's disease today, early diagnosis is important to slow down the adverse conditions that may arise and to improve the quality of life. As a result of the development of artificial intelligence technologies and their consistent application in different fields, machine learning techniques have the potential to play an important role in the detection of Alzheimer's disease. In particular, deep learning-based methods, which have the ability to automatically extract patterns from complex patterns, are promising in this field. Recent studies show that the use of deep learning models for Alzheimer's detection on images is becoming widespread. In addition to contributing to the early diagnosis of the disease, these models also show potential in detecting different stages of the disease by analyzing the symptoms in magnetic resonance images. These developments enable the development of more effective treatment methods for patients. However, more studies are needed to evaluate the efficacy and safety of these technologies in clinical applications. In this study, classification studies were performed using MobileNetV2, InceptionV3, Xception, Vgg16 and Vgg19 models for the diagnosis of the disease on a publicly shared Alzheimer's dataset consisting of 6400 different samples and 4 different classes. An accuracy of 99.92% was calculated for the MobileNetV2 model. The performances of the models used in this study were compared with similar studies in the literature and their performances were reported in terms of different metrics. Among the five different models used, the highest accuracy value of 99.92% was obtained with MobileNetV2. It was concluded that the architectures used in the experimental studies produced generally better results than similar studies in the literature.

High-Precision Detection of Lung Adenocarcinoma Using Augmented VGG16 and Transfer Learning

Objectives: The research aims to enhance the effectiveness of the deep learning model for detecting lung adenocarcinoma using VGG16 CNN with transfer learning on medical images. The proposed model addresses the overfitting issues of existing models, thereby improving lung cancer detection accuracy. Methods: The lung CT scan images are analyzed for detection and classification using VGG16, leveraging transfer learning techniques. The dataset includes labeled lung images from LUNA 16, Kaggle, and other datasets, which are augmented to introduce variability and reduce overfitting. The images are resized format to 224 × 224 pixels. During the validation and training phases, the model's performance is evaluated based on accuracy, and precision. Findings: The adjusted VGG16 model attains a training accuracy of 99.42%, a validation accuracy of 99.13%, and a validation precision of 95.45%, indicating strong generalization capabilities, outperforming other existing models. Novelty: This research presents a new method for detecting lung cancer by combining advanced preprocessing and data augmentation techniques with the VGG16 architecture using transfer learning. This combination greatly enhances the model's ability to detect cancer accurately, setting a new standard in the field. Keywords: Lung cancer, VGG16, Transfer learning, Data augmentation, medical imaging, CT (Computer Tomography)

Predicting an opaque bubble layer during small-incision lenticule extraction surgery based on deep learning.

This study aimed to predict the formation of OBL during femtosecond laser SMILE surgery by employing deep learning technology. This was a cross-sectional, retrospective study conducted at a university hospital. Surgical videos were randomly divided into a training (3,271 patches, 73.64%), validation (704 patches, 15.85%), and internal verification set (467 patches, 10.51%). An artificial intelligence (AI) model was developed using a SENet-based residual regression deep neural network. Model performance was assessed using the mean absolute error (E MA ), Pearson's correlation coefficient (r), and determination coefficient (R 2 ). Four distinct types of deep neural network models were established. The modified deep residual neural network prediction model with channel attention built on the PyTorch framework demonstrated the best predictive performance. The predicted OBL area values correlated well with the Photoshop-based measurements (E MA = 0.253, r = 0.831, R 2 = 0.676). The ResNet (E MA = 0.259, r = 0.798, R 2 = 0.631) and Vgg19 models (E MA = 0.31, r = 0.758, R 2 = 0.559) both displayed satisfactory predictive performance, while the U-net model (E MA = 0.605, r = 0.331, R 2 = 0.171) performed poorest. We used a panoramic corneal image obtained before the SMILE laser scan to create a unique deep residual neural network prediction model to predict OBL formation during SMILE surgery. This model demonstrated exceptional predictive power, suggesting its clinical applicability across a broad field.

Development of an Artificial Intelligence System for Distinguishing Malignant from Benign Soft Tissue Tumors Using Contrast-Enhanced MR Images.

The integration of artificial intelligence (AI) into orthopedics has enhanced the diagnosis of various conditions; however, its use in diagnosing soft-tissue tumors remains limited owing to its complexity. This study aimed to develop and assess an AI-driven diagnostic support system for magnetic resonance imaging (MRI)-based soft-tissue tumor diagnosis, potentially improving accuracy and aiding radiologists and orthopedic surgeons. Experienced orthopedic oncologists and radiologists annotated 720 images from 77 cases (41 benign and 36 malignant soft-tissue tumors). Eleven tumor subtypes were identified and classified into benign and malignant groups based on histological diagnosis. Utilizing the standard machine learning classifier pipeline, we examined and down-selected imaging protocols and their predominant radiomic features within the tumor's three-dimensional region to differentiate between benign and malignant tumors. Among the scan protocols, contrast-enhanced T1 weighted fat-suppressed images showed the most accurate classification based on radiomics features. We focused on the two-dimensional features from the largest tumor boundary surface and its neighboring slices, leveraging texture-based radiomic and deep convolutional neural network features from a pretrained VGG19 model. The test data comprised 44 contrast-enhanced images (22 benign and 22 malignant soft-tissue tumors) containing six malignant and five benign subtypes distinct from the training data. We compared expert and non-expert human performances against AI by assessing malignancy detection and the time required for classification. The AI model showed comparable accuracy (AUC 0.91) to that of radiologists (AUC 0.83) and orthopedic surgeons (AUC 0.73). Notably, the AI model processed data approximately 400 times faster than its human counterparts, showcasing its capacity to significantly boost diagnostic efficiency. We developed an AI-driven diagnostic support system for MRI-based soft-tissue tumor diagnosis. While additional refinement is necessary for clinical applications, our system has exhibited promising potential in differentiating between benign and malignant soft-tissue tumors based on MRI.

Defect Detection of Polyethylene Gas Pipeline Based on Convolutional Neural Networks and Image Processing

Abstract In this paper, a method based on image recognition was proposed to detect the defects of polyethylene (PE) gas pipeline, especially the deformation due to the indentation. First, the pipeline -detection VGG (PD-VGG) model was established based on the convolutional neural network (CNN), and appropriate model parameters were optimized through model training. The defect recognition rate of the improved model can reach 94.76%. Following, the weighted average graying algorithm was used to separate the defects characterized by deformation. Then, an improved gamma correction algorithm was applied to achieve image enhancement, and the interference of impurities adhered on intersurface of pipeline was also removed by using multilayer filters. The edge detection of the defect image was completed by using the Canny operator, and following the screening between the target contour and the interference contour by using top-contour. Finally, the algorithm for minimum outer rectangle algorithm was used to fit the defect contour, and the eigenvalues of deformation defects were extracted. The results indicate that the above defect detection method can better extract the deformation contour of the dented pipeline. The high agreement with the experimental results provides a basis for the research of effectively recognizing whether the pipeline has undergone ductile failure only through profile detection of defects.

Unlocking the potential of simulated hyperspectral imaging in agro environmental analysis: a comprehensive study of algorithmic approaches

This study focuses on identifying and evaluating the severity of powdery mildew disease in tomato plants. The uniqueness of this work lies in combining the imaging and advanced deep learning methods to develop a technique that transforms Red Green Blue (RGB) images into Simulated Hyperspectral Images (SHSI) to perform spectral and spatial analysis for precise detection and assessment of powdery mildew severity, thereby enhancing disease management. Furthermore, this research evaluates three advanced pre-trained VGG16 models, ResNet50 and EfficientNet-B7 algorithms for image preprocessing and feature extraction. Extracted features are passed to a neural network generator model to convert RGB image features into SHSIs, providing insights into the spectrum. This method enables the image analysis to perform assessments from SHSIs for health classification using Normalized Difference Vegetation Index (NDVI) values, which are meticulously compared with accurate hyperspectral data using metrics like mean absolute error (MAE) and root mean squared error (RMSE). This strategy enhances precision farming, environmental monitoring, and remote sensing accuracy. Results show that ResNet50’s architecture offers a robust framework for this study’s spectral and spatial analysis, making it a suitable choice over VGG16 and EfficientNet-B7 for transforming RGB images into SHSI. These simulated hyperspectral images offer a scalable and affordable approach for precise assessment of crop disease severity.

Auto encoder-based defense mechanism against popular adversarial attacks in deep learning.

Convolutional Neural Network (CNN)-based models are prone to adversarial attacks, which present a significant hurdle to their reliability and robustness. The vulnerability of CNN-based models may be exploited by attackers to launch cyber-attacks. An attacker typically adds small, carefully crafted perturbations to original medical images. When a CNN-based model receives the perturbed medical image as input, it misclassifies the image, even though the added perturbation is often imperceptible to the human eye. The emergence of such attacks has raised security concerns regarding the implementation of deep learning-based medical image classification systems within clinical environments. To address this issue, a reliable defense mechanism is required to detect adversarial attacks on medical images. This study will focus on the robust detection of pneumonia in chest X-ray images through CNN-based models. Various adversarial attacks and defense strategies will be evaluated and analyzed in the context of CNN-based pneumonia detection. From earlier studies, it has been observed that a single defense mechanism is usually not effective against more than one type of adversarial attack. Therefore, this study will propose a defense mechanism that is effective against multiple attack types. A reliable defense framework for pneumonia detection models will ensure secure clinical deployment, facilitating radiologists and doctors in their diagnosis and treatment planning. It can also save time and money by automating routine tasks. The proposed defense mechanism includes a convolutional autoencoder to denoise perturbed Fast Gradient Sign Method (FGSM) and Projected Gradient Descent (PGD) adversarial images, two state- of-the-art attacks carried out at five magnitudes, i.e., ε (epsilon) values. Two pre-trained models, VGG19 and VGG16, and our hybrid model of MobileNetV2 and DenseNet169, called Stack Model, have been used to compare their results. This study shows that the proposed defense mechanism outperforms state-of-the-art studies. The PGD attack using the VGG16 model shows a better attack success rate by reducing overall accuracy by up to 67%. The autoencoder improves accuracy by up to 16% against PGD attacks in both the VGG16 and VGG19 models.

Pruning convolution neural networks using filter clustering based on normalized cross-correlation similarity

ABSTRACT Despite all the recent development and success of deep neural networks, deployment of a deep model onto the resource-constrained devices still remains challenging. However, model pruning can resolve this issue for Convolutional Neural Networks (CNNs), since it is one of the most popular approaches to reducing computational complexities. Therefore, this article presents a pruning model for convolutional neural networks. The proposed method classifies and arranges similar filters into the same cluster where the similarity is calculated using a three-dimensional normalized cross-correlation. Moreover, these steps can be completed entirely based on the filter values while not requiring a set of test images as well as the acquisition of any filter activation. In the research, the performances of the proposed model pruning method have been evaluated, where it is observed that the proposed approach is computationally light and requires significantly less time and resources compared to ML and activation-based approaches. In the experiments, using the VGG16 model on the Cifar10 dataset, the proposed approach results in the pruned model(s) which are comparable in performance with models found using activation-based methods and expensive ML-based methods. Similar results are found when pruning a custom CNN on the MNIST and Fashion MNIST datasets as well.