Pre-trained Deep Convolutional Neural Network Research Articles

Hybrid Deep Neural Network for Facial Expressions Recognition

Facial expressions are critical indicators of human emotions where recognizing facial expressions has captured the attention of many academics, and recognition of expressions in natural situations remains a challenge due to differences in head position, occlusion, and illumination. Several studies have focused on recognizing emotions from frontal images only, while in this paper wild images from the FER2013 dataset have been used to make a more generalizing model with the existence of its challenges, it is among the most difficult datasets that only got 65.5 % accuracy human-level. This paper proposed a model for recognizing facial expressions using pre-trained deep convolutional neural networks and the technique of transfer learning. this hybrid model used a combination of two pre-trained deep convolutional neural networks, training the model in multiple cases for more efficiency to categorize the facial expressions into seven classes. The results show that the best accuracy of the suggested models is 74.39% for the hybrid model, and 73.33% for Fine-tuned the single EfficientNetB0 model, while the highest accuracy for previous methods was 73.28%. Thus, the hybrid and single models outperform other state of art classification methods without using any additional, the hybrid and single models ranked in the first and second position among these methods. Also, The hybrid model has even outperformed the second-highest in accuracy method which used extra data. The incorrectly labeled images in the dataset unfairly reduce accuracy but our best model recognized their actual classes correctly.

Efficient structural pseudoinverse learning-based hierarchical representation learning for skin lesion classification

The success of deep learning in skin lesion classification mainly depends on the ultra-deep neural network and the significantly large training data set. Deep learning training is usually time-consuming, and large datasets with labels are hard to obtain, especially skin lesion images. Although pre-training and data augmentation can alleviate these issues, there are still some problems: (1) the data domain is not consistent, resulting in the slow convergence; and (2) low robustness to confusing skin lesions. To solve these problems, we propose an efficient structural pseudoinverse learning-based hierarchical representation learning method. Preliminary feature extraction, shallow network feature extraction and deep learning feature extraction are carried out respectively before the classification of skin lesion images. Gabor filter and pre-trained deep convolutional neural network are used for preliminary feature extraction. The structural pseudoinverse learning (S-PIL) algorithm is used to extract the shallow features. Then, S-PIL preliminarily identifies the skin lesion images that are difficult to be classified to form a new training set for deep learning feature extraction. Through the hierarchical representation learning, we analyze the features of skin lesion images layer by layer to improve the final classification. Our method not only avoid the slow convergence caused by inconsistency of data domain but also enhances the training of confusing examples. Without using additional data, our approach outperforms existing methods in the ISIC 2017 and ISIC 2018 datasets.

Optimizing the transfer-learning with pretrained deep convolutional neural networks for first stage breast tumor diagnosis using breast ultrasound visual images.

Female accounts for approximately 50% of the total population worldwide and many of them had breast cancer. Computer-aided diagnosis frameworks could reduce the number of needless biopsies and the workload of radiologists. This research aims to detect benign and malignant tumors automatically using breast ultrasound (BUS) images. Accordingly, two pretrained deep convolutional neural network (CNN) models were employed for transfer learning using BUS images like AlexNet and DenseNet201. A total of 697 BUS images containing benign and malignant tumors are preprocessed and performed classification tasks using the transfer learning-based CNN models. The classification accuracy of the benign and malignant tasks is completed and achieved 92.8% accuracy using the DensNet201 model. The results thus achieved compared in state of the art using benchmark data set and concluded proposed model outperforms in accuracy from first stage breast tumor diagnosis. Finally, the proposed model could help radiologists diagnose benign and malignant tumors swiftly by screening suspected patients.

A convolutional deep learning model for improving mammographic breast-microcalcification diagnosis

This study aimed to assess the diagnostic performance of deep convolutional neural networks (DCNNs) in classifying breast microcalcification in screening mammograms. To this end, 1579 mammographic images were collected retrospectively from patients exhibiting suspicious microcalcification in screening mammograms between July 2007 and December 2019. Five pre-trained DCNN models and an ensemble model were used to classify the microcalcifications as either malignant or benign. Approximately one million images from the ImageNet database had been used to train the five DCNN models. Herein, 1121 mammographic images were used for individual model fine-tuning, 198 for validation, and 260 for testing. Gradient-weighted class activation mapping (Grad-CAM) was used to confirm the validity of the DCNN models in highlighting the microcalcification regions most critical for determining the final class. The ensemble model yielded the best AUC (0.856). The DenseNet-201 model achieved the best sensitivity (82.47%) and negative predictive value (NPV; 86.92%). The ResNet-101 model yielded the best accuracy (81.54%), specificity (91.41%), and positive predictive value (PPV; 81.82%). The high PPV and specificity achieved by the ResNet-101 model, in particular, demonstrated the model effectiveness in microcalcification diagnosis, which, in turn, may considerably help reduce unnecessary biopsies.

A Novel Technique for Classifying Bird Damage to Rapeseed Plants Based on a Deep Learning Algorithm

Estimation of crop damage plays a vital role in the management of fields in the agriculture sector. An accurate measure of it provides key guidance to support agricultural decision-making systems. The objective of the study was to propose a novel technique for classifying damaged crops based on a state-of-the-art deep learning algorithm. To this end, a dataset of rapeseed field images was gathered from the field after birds’ attacks. The dataset consisted of three classes including undamaged, partially damaged, and fully damaged crops. Vgg16 and Res-Net50 as pre-trained deep convolutional neural networks were used to classify these classes. The overall classification accuracy reached 93.7% and 98.2% for the Vgg16 and the ResNet50 algorithms, respectively. The results indicated that a deep neural network has a high ability in distinguishing and categorizing different image-based datasets of rapeseed. The findings also revealed a great potential of deep learning-based models to classify other damaged crops.

Convolutional Neural Network for Skull Recognition

Automatic skull identification systems play a vital role for forensic law authorities to recognize victim identity. Motivated by potential applications of these kinds of systems, this research aims to apply a pre-trained deep convolutional neural network (CNN) for face skull recognition. Basically, the unknown skull image is fed to a pre-trained CNN network to extract a 1D feature vector, and then it will be matched with photos at database agencies to identify the closest match. To validate the proposed skull recognition system, it has been applied for a total of 13 skulls, and the reported results indicated a good was achieved. In addition, various CNN architectures were investigated, including shallow, medium, and deep CNN models. The best performance was reported from the shallow CNN model with a 92% recognition rate.

Efficient masked face recognition method during the COVID-19 pandemic.

The coronavirus disease (COVID-19) is an unparalleled crisis leading to a huge number of casualties and security problems. In order to reduce the spread of coronavirus, people often wear masks to protect themselves. This makes face recognition a very difficult task since certain parts of the face are hidden. A primary focus of researchers during the ongoing coronavirus pandemic is to come up with suggestions to handle this problem through rapid and efficient solutions. In this paper, we propose a reliable method based on occlusion removal and deep learning-based features in order to address the problem of the masked face recognition process. The first step is to remove the masked face region. Next, we apply three pre-trained deep Convolutional Neural Networks (CNN), namely VGG-16, AlexNet, and ResNet-50, and use them to extract deep features from the obtained regions (mostly eyes and forehead regions). The Bag-of-features paradigm is then applied to the feature maps of the last convolutional layer in order to quantize them and to get a slight representation comparing to the fully connected layer of classical CNN. Finally, Multilayer Perceptron (MLP) is applied for the classification process. Experimental results on Real-World-Masked-Face-Dataset show high recognition performance compared to other state-of-the-art methods.

An empirical study on temporal modeling for online action detection

Online action detection (OAD) is a practical yet challenging task, which has attracted increasing attention in recent years. A typical OAD system mainly consists of three modules: a frame-level feature extractor which is usually based on pre-trained deep Convolutional Neural Networks (CNNs), a temporal modeling module, and an action classifier. Among them, the temporal modeling module is crucial which aggregates discriminative information from historical and current features. Though many temporal modeling methods have been developed for OAD and other topics, their effects are lack of investigation on OAD fairly. This paper aims to provide an empirical study on temporal modeling for OAD including four meta types of temporal modeling methods, i.e. temporal pooling, temporal convolution, recurrent neural networks, and temporal attention, and uncover some good practices to produce a state-of-the-art OAD system. Many of them are explored in OAD for the first time, and extensively evaluated with various hyper parameters. Furthermore, based on our empirical study, we present several hybrid temporal modeling methods. Our best networks, i.e. , the hybridization of DCC, LSTM and M-NL, and the hybridization of DCC and M-NL, which outperform previously published results with sizable margins on THUMOS-14 dataset (48.6% vs. 47.2%) and TVSeries dataset (84.3% vs. 83.7%).

Application Value of a Deep Convolutional Neural Network Model for Cytological Assessment of Thyroid Nodules.

Objective To investigate the application value of a deep convolutional neural network (CNN) model for cytological assessment of thyroid nodules. Methods 117 patients with thyroid nodules who underwent thyroid cytology examination in the Affiliated People's Hospital of Ningbo University between January 2017 and December 2019 were included in this study. 100 papillary thyroid cancer samples and 100 nonmalignant samples were collected respectively. The sample images were translated vertically and horizontally. Thus, 900 images were separately created in the vertical and horizontal directions. The sample images were randomly divided into training samples (n = 1260) and test samples (n = 540) at the ratio of 7 : 3 per the training sample to test sample. According to the training samples, the pretrained deep convolutional neural network architecture Resnet50 was trained and fine-tuned. A convolutional neural network-based computer-aided detection (CNN-CAD) system was constructed to perform full-length scan of the test sample slices. The ability of CNN-CAD to screen malignant tumors was analyzed using the threshold setting method. Eighty pathological images were collected from patients who received treatment between January 2020 and May 2020 and used to verify the value of CNN in the screening of malignant thyroid nodules as verification set. Results With the number of iterations increasing, the training and verification loss of CNN model gradually decreased and tended to be stable, and the training and verification accuracy of CNN model gradually increased and tended to be stable. The average loss rate of training samples determined by the CNN model was (22.35 ± 0.62) %, and the average loss rate of test samples determined by the CNN model was (26.41 ± 3.37) %. The average accuracy rate of training samples determined by the CNN model was (91.04 ± 2.11) %, and the average accuracy rate of test samples determined by the CNN model was (91.26 ± 1.02)%. Conclusion A CNN model exhibits a high value in the cytological diagnosis of thyroid diseases which can be used for the cytological diagnosis of malignant thyroid tumor in the clinic.

Detection of invisible cracks in ceramic materials using by pre-trained deep convolutional neural network

Ceramic materials are an indispensable part of our lives. Today, ceramic materials are mainly used in construction and kitchenware production. The fact that some deformations cannot be seen with the naked eye in the ceramic industry leads to a loss of time in the detection of deformations in the products. Delays that may occur in the elimination of deformations and in the planning of the production process cause the products with deformation to be excessive, which adversely affects the quality. In this study, a deep learning model based on acoustic noise data and transfer learning techniques was designed to detect cracks in ceramic plates. In order to create a data set, noise curves were obtained by applying the same magnitude impact to the ceramic experiment plates by impact pendulum. For experimental application, ceramic plates with three invisible cracks and one undamaged ceramic plate were used. The deep learning model was trained and tested for crack detection in ceramic plates by the data set obtained from the noise graphs. As a result, 99.50% accuracy was achieved with the deep learning model based on acoustic noise.

Comparative Experimental Investigation of Deep Convolutional Neural Networks for Latent Fingerprint Pattern Classification

Fingerprint pattern recognition is of great importance in forensic examinations and in helping diagnose some diseases. The automatic realization of fingerprint recognition processes can take time due to the feature extraction process in classical machine learning or deep learning methods. In this study, the effective use of deep convolutional neural networks (DCNN) in fingerprint pattern recognition and classification, in which feature extraction takes place automatically, was examined experimentally and comparatively. Five DCNN models have been designed and implemented with a transfer learning approach. Four of these five models are Alexnet, Googlenet, Resnet-18, and Squeezenet pre-trained DCNN models. The fifth model is the DCNN model designed from the ground up. It was concluded that the designed DCNN models can be used effectively in fingerprint recognition and classification, and that fast results can be obtained and generalized with advanced DCNN models.

Superiority of Supervised Machine Learning on Reading Chest X-Rays in Intensive Care Units.

Purpose: Portable chest radiographs are diagnostically indispensable in intensive care units (ICU). This study aimed to determine if the proposed machine learning technique increased in accuracy as the number of radiograph readings increased and if it was accurate in a clinical setting.Methods: Two independent data sets of portable chest radiographs (n = 380, a single Japanese hospital; n = 1,720, The National Institution of Health [NIH] ChestX-ray8 dataset) were analyzed. Each data set was divided training data and study data. Images were classified as atelectasis, pleural effusion, pneumonia, or no emergency. DenseNet-121, as a pre-trained deep convolutional neural network was used and ensemble learning was performed on the best-performing algorithms. Diagnostic accuracy and processing time were compared to those of ICU physicians.Results: In the single Japanese hospital data, the area under the curve (AUC) of diagnostic accuracy was 0.768. The area under the curve (AUC) of diagnostic accuracy significantly improved as the number of radiograph readings increased from 25 to 100% in the NIH data set. The AUC was higher than 0.9 for all categories toward the end of training with a large sample size. The time to complete 53 radiographs by machine learning was 70 times faster than the time taken by ICU physicians (9.66 s vs. 12 min). The diagnostic accuracy was higher by machine learning than by ICU physicians in most categories (atelectasis, AUC 0.744 vs. 0.555, P < 0.05; pleural effusion, 0.856 vs. 0.706, P < 0.01; pneumonia, 0.720 vs. 0.744, P = 0.88; no emergency, 0.751 vs. 0.698, P = 0.47).Conclusions: We developed an automatic detection system for portable chest radiographs in ICU setting; its performance was superior and quite faster than ICU physicians.

Automated detection of leukemia by pretrained deep neural networks and transfer learning: A comparison

Leukemia is usually diagnosed by viewing the smears of blood and bone marrow using microscopes and complex Cytochemical tests can be used to authorize and classify leukemia. But these methods are costly, slow and affected by the proficiency and expertise of the specialists concerned. Leukemia can be detected with the help of image processing-based methods by analyzing microscopic smear images to detect the presence of leukemic cells and such techniques are simple, fast, cheap and not biased by the specialists. The proposed study presents a computer aided diagnosis system that uses pretrained deep Convolutional Neural Networks (CNNs) for detection of leukemia images against normal images. The use of pretrained networks is comparatively an easy method of applying deep learning for image analysis and the comparison results of the present study can be used to choose appropriate networks for diagnostic tasks. The microscopic images used in the proposed work were downloaded from a public dataset ALL-IDB. In the proposed work, image classification is done without using any image segregation and feature extraction practices and the study used pretrained series network AlexNet, VGG-16, VGG-19, Directed Acyclic Graph (DAG) networks GoogLeNet, Inceptionv3, MobileNet-v2, Xception, DenseNet-201, Inception-ResNet-v2 and residual networks ResNet-18, ResNet-50 and ResNet-101 for performing the classification and comparison. A classification accuracy of 100% is obtained with all the pretrained networks used in the study for ALL_IDB1 dataset and for ALL_IDB2 dataset, 100% accuracy is obtained with all networks except the AlexNet and VGG-16. The efficacy of three optimization algorithms Stochastic Gradient Descent with Momentum (SGDM), Root Mean Square propagation (RMSprop) and Adaptive Moment estimation (ADAM) is also compared in all the classifications performed. The study considered the detection of leukemia in general only, and classification of leukemia into different types can be attempted as a future work.

A Joint Optimization Framework of the Embedding Model and Classifier for Meta-Learning

The aim of meta-learning is to train the machine to learn quickly and accurately. Improving the performance of the meta-learning model is important in solving the problem of small samples and in achieving general artificial intelligence. A meta-learning method based on feature embedding that exhibits good performance on the few-shot problem was previously proposed. In this method, the pretrained deep convolution neural network was used as the embedding model of sample features, and the output of one layer was used as the feature representation of samples. The main limitation of the method is the inability to fuse low-level texture features and high-level semantic features of the embedding model and joint optimization of the embedding model and classifier. Therefore, a multilayer adaptive joint training and optimization method of the embedding model was proposed in the current study. The main characteristics of the current method include using multilayer adaptive hierarchical loss to train the embedding model and using the quantum genetic algorithm to jointly optimize the embedding model and classifier. Validation was performed based on multiple public datasets for meta-learning model testing. The proposed method shows higher accuracy compared with multiple baseline methods.

A Simplified Framework for the Detection of Intracranial Hemorrhage in CT Brain Images Using Deep Learning.

The need for accurate and timely detection of Intracranial hemorrhage (ICH) is of utmost importance to avoid untoward incidents that may even lead to death. Hence, this presented work leverages the ability of a pretrained deep convolutional neural network (CNN) for the detection of ICH in computed tomography (CT) brain images. Different frameworks have been analyzed for their effectiveness for the classification of CT brain images into hemorrhage or non-hemorrhage conditions. All these frameworks were investigated on the CQ500 dataset. Furthermore, an exclusive preprocessing pipeline was designed for both normal and ICH CT images. Firstly, a framework involving the pretrained deep CNN, AlexNet, has been exploited for both feature extraction and classification using the transfer learning method. Secondly, a modified AlexNet-Support vector machine (SVM) classifier is explored, and finally, a feature selection method, Principal Component Analysis (PCA), has been introduced in the AlexNet-SVM classifier model, and its efficacy is also explored. These models were trained and tested on two different sets of CT images, one containing the original images without preprocessing and another set consisting of preprocessed images. The modified AlexNet-SVM classifier has shown an improved performance in comparison to the other investigated frameworks and has achieved a classification accuracy of 99.86% and sensitivity and specificity of 0.9986 for the detection of ICH in the brain CT images. This research has given an overview of a simple and efficient framework for the classification of hemorrhage and non-hemorrhage images. The proposed simplified deep learning framework also manifests its ability as a screening tool to assist the radiological trainees in the accurate detection of ICH.

Efficient Cloaked Face Recognition Methodology throughout The Covid-19 Pandemic

The COVID-19 is a partner in Nursing unequaled disaster resulting in a huge range of casualties and protection issues. to cut back the unfold of coronavirus, individuals typically wear masks to guard themselves. This makes face popularity a truly tough project because bound components of the face rectangular measure hidden. A primary awareness of researchers for the duration of the continuing coronavirus pandemic is to come back up with hints to handle this downside thru fast and reasonably-priced solutions. during this paper, we tend to endorse a dependable technique supported by discard cloaked region and deep learning-based options to deal with the matter of the cloaked face recognition technique. the number one step is to discard the cloaked face vicinity. next, we tend to apply pre-trained deep Convolutional neural networks (CNN) to extract the only options from the received areas (in general eyes and forehead regions). in the end, the Bag-of-features paradigm is carried out on the function maps of the last convolutional layer to quantize them and to induce small illustration scrutiny to the simply related layer of classical CNN. in the end, Multilayer Perceptron (MLP) is implemented for the class approach. Experimental effects on real-global-Masked-Face-Dataset display high popularity overall performance.

A two-stage deep transfer learning for localisation of forced oscillations disturbance source

Accurately locating forced oscillations (FOs) disturbance source in a large-scale power system is a challenging task. In this paper, a localisation method featuring system-level and area-level localisation is proposed. Compared with traditional localisation methods which process the FOs signals only in control centre phasor data concentrator (PDC), the proposed method can not only relieve the data communication pressure, but also meets the data privacy and confidentiality requirements of utility companies. Besides, with the two-stage deep transfer learning, high accuracy of localisation can be realised with far less training. Firstly, by adopting principle component analysis (PCA) to extract the most representative information at system-level and smooth pseudo Wigner-Ville distribution (SPWVD) to characterize FOs signals, graphical representations of FOs at system-level and area-level are obtained respectively, transforming the localisation problem to image recognition problem. Subsequently, a two-stage deep transfer learning (DTL) algorithm is developed to locate FOs disturbance source. The first stage involves repurposing the learnt features from a pre-trained deep convolutional neural network (previously trained for universal image recognition) to improve the learning of system-level localisation. In a similar logic, the second stage entails transferring the knowledge acquired from the first stage is exploited to aid area-level localisation learning. Case studies carried out on the WECC 179-bus system demonstrate that the proposed method achieves a significantly higher accuracy in the presence of measurement noise, topology variation and load disturbances of the system with respect to the traditional machine learning method. The proposed method also exhibits a more favorable computational performance and learning efficiency due to the employment of the tactfully designed two-stage DTL approach.

A Cautionary Note on Predicting Social Judgments from Faces with Deep Neural Networks

People spontaneously infer other people’s psychology from faces, encompassing inferences of their affective states, cognitive states, and stable traits such as personality. These judgments are known to be often invalid, but nonetheless bias many social decisions. Their importance and ubiquity have made them popular targets for automated prediction using deep convolutional neural networks (DCNNs). Here, we investigated the applicability of this approach: how well does it generalize, and what biases does it introduce? We compared three distinct sets of features (from a face identification DCNN, an object recognition DCNN, and using facial geometry), and tested their prediction across multiple out-of-sample datasets. Across judgments and datasets, features from both pre-trained DCNNs provided better predictions than did facial geometry. However, predictions using object recognition DCNN features were not robust to superficial cues (e.g., color and hair style). Importantly, predictions using face identification DCNN features were not specific: models trained to predict one social judgment (e.g., trustworthiness) also significantly predicted other social judgments (e.g., femininity and criminal), and at an even higher accuracy in some cases than predicting the judgment of interest (e.g., trustworthiness). Models trained to predict affective states (e.g., happy) also significantly predicted judgments of stable traits (e.g., sociable), and vice versa. Our analysis pipeline not only provides a flexible and efficient framework for predicting affective and social judgments from faces but also highlights the dangers of such automated predictions: correlated but unintended judgments can drive the predictions of the intended judgments.

Machine Learning with Remote Sensing Image Data Sets

Computer vision, as a part of machine learning, gains significant attention from researches nowadays. Aerial scene classification is a prominent chapter of computer vision with a vast application: military, surveillance and security, environment monitoring, detection of geospatial objects, etc. There are several publicly available remote sensing image datasets, which enable the deployment of various aerial scene classification algorithms. In our article, we use transfer learning from pre-trained deep Convolutional Neural Networks (CNN) within remote sensing image classification. Neural networks utilized in our research are high-dimensional previously trained CNN on ImageNet dataset. Transfer learning can be performed through feature extraction or fine-tuning. We proposed a two-stream feature extraction method and afterward image classification through a handcrafted classifier. Fine-tuning was performed with adaptive learning rates and a regularization method label smoothing. The proposed transfer learning techniques were validated on two remote sensing image datasets: WHU RS datasets and AID dataset. Our proposed method obtained competitive results compared to state-of-the-art methods.

Early Detection of Mold-Contaminated Peanuts Using Machine Learning and Deep Features Based on Optical Coherence Tomography

Fungal infection is a pre-harvest and post-harvest crisis for farmers of peanuts. In environments with temperatures around 28 °C to 30 °C or relative humidity of approximately 90%, mold-contaminated peanuts have a considerable likelihood to be infected with Aflatoxins. Aflatoxins are known to be highly carcinogenic, posing danger to humans and livestock. In this work, we proposed a new approach for detection of mold-contaminated peanuts at an early stage. The approach employs the optical coherence tomography (OCT) imaging technique and an error-correcting output code (ECOC) based Support Vector Machine (SVM) trained on features extracted using a pre-trained Deep Convolutional Neural Network (DCNN). To this end, mold-contaminated and uncontaminated peanuts were scanned to create a data set of OCT images used for training and evaluation of the ECOC-SVM model. Results showed that the proposed approach is capable of detecting mold-contaminated peanuts with respective accuracies of approximately 85% and 96% after incubation periods of 48 and 96 h.