Convolutional Neural Network

Detecting occluded faces is a non-trivial problem for face detection in computer vision. This challenge becomes more difficult when the occlusion covers majority of the face. Despite the high performance of current state-of-the-art face detection algorithms, the detection of occluded and covered faces is an unsolved problem and is still worthy of study. In this paper, a deep-learning-face-detection model Niqab-Face-Detector is proposed along with context-based labeling technique for detecting unconstrained veiled faces such as faces covered with niqab. An experimental test was conducted to evaluate the performances of the proposed model using the Niqab-Face dataset. The experiment showed encouraging results and improved accuracy compared with state-of-the-art face detection algorithms

Automating the processing of videos in applications such as surveillance, sport commentary and activity detection, human-machine interaction, and health/disability care is crucial to their correct functioning. In such video processing tasks, recognition of various human actions is a pivotal component for the correct understanding of videos and making decisions upon it. Accurately recognizing human actions is a complex process, demanding high computing capabilities and intelligent algorithms. Several factors, such as object occlusion, camera movement, and background clutter, further challenge the task and its accuracy, essentially leaving deep learning approaches the only viable option for properly detecting human actions in videos. In this study, we propose CoReHAR, a novel Human Action Recognition method that employs both deep Convolutional and Recurrent neural networks on raw video frames. Using the pre-trained ResNet152 CNN, deep features are initially extracted from video frames. The sequential information of the frames is then learned using DB-LSTM RNN. Multiple stacked layers in forward and backward passes of the DB-LSTM provide increased network depth for higher accuracy. A number of techniques are also applied to improve CoReHAR’s processing speed on heterogeneous GPU-enabled systems. The proposed method is evaluated using PyTorch, and is compared to the state-of-the-art methods, showing a considerable efficiency increase, with nearly 95% recognition accuracy measured as an average over all splits of the challenging UCF101 dataset.

The human chest contains vital organs such as the heart, lungs, and other organs. Chest radiology is one of the best and least costly methods to diagnose chest diseases. In this study, proposed a new method to diagnose 14 main diseases of the chest such as (cardiomegaly, emphysema, effusion, hernia, nodule, pneumothorax, atelectasis, pleural - thickening, mass, edema, integration, penetration, fibrosis, pneumonia) using the neural network and deep learning to increase accuracy, sensitivity, and specificity. The proposed method is implemented in the form of a web application and is available as a decision-making system for physicians to diagnose chest diseases.The results of the simulation on the sample dataset showed that the diagnosis of chest diseases was 98.93%, indicating the high efficiency of the new method. Finally, the proposed method was compared with other deep learning architectures such as densenet121, vgg16, exception architecture on the same dataset, which showed a 5% higher accuracy than them.

Review spam is an opinion written to promote or demote a product or brand on websites and other internet services by some users. Since it is not easy for humans to recognize these types of opinions, a model can be provided to detect them. In recent years, much research has been done to detect these types of reviews, and with the expansion of deep neural networks and the efficiency of these networks in various issues, in recent years, multiple types of deep neural networks have been used to identify spam reviews. This paper reviews the proposed deep learning methods for the problem of review spam detection. Challenges, evaluation criteria, and datasets in this area are also examined.

Spams are well-known examples of unsolicited text or messages which are sent by unknown individuals and cause issues for smartphone users. The inconvenience imposed on users, the loss of network traffic, the rise in the calculated cost, occupying more physical space on the mobile phone, and abusing and defrauding recipients are but a few of their downsides. Consequently, the automated identification of  suspicious and spam messages is undoubtedly vitally important. Additionally, text messages which are smartly composed might be difficult to recognize. However, the present methodologies in this subject are hindered by the absence of adequate Persian datasets. A huge body of research and experiments has revealed that techniques based on deep and combined learning are superior at identifying unpleasant text messages. This work sought to develop an effective strategy for identifying SMS spam through utilizing combining machine learning classification algorithms together with deep learning models. After applying  preprocessing on our gathered dataset, the suggested technique applies two convolutional neural network layers, the first of which being an LSTM layer, and the second one which is a fully connected layer to extract the data characteristics, thereby implementing the suggested deep learning approach. As part of the Machine Learning methodologies, the vector support machine makes use of the data and features at hand to determine the ultimate classification. Results indicate that the suggested model is implemented more effectively than the existing techniques, and an accuracy of 97.7% was achieved as a result.

In recent scenario, women are suffering from breast cancer disease across the world. Mammography is one of the important methods to detect breast cancer early; that to reduce the cost and workload of radiologists. Medical image processing is a tremendous technique used to determine the disease in advance to reduce the risk factor. To predict the disease from 2-D mammography images for diagnosing and detecting based on advanced soft computing paradigm. Still, to get more accuracy in all coordinate axes, 3-D mammography imaging is used to capture depth information from all different angles. After the reconstruction of this process, a better quality of 3D mammography is obtained. It is useful for the experts to identify the disease in well advance. To improve the accuracy of disease findings, deep convolution neural networks (CNN) can be applied for automatic feature learning, and classifier building. This work also presents a comparison of the other state of art methods used in the last decades.

Lung infection is the most dangerous sign of Covid 19. X-ray images are the most effective means of diagnosing this virus. In order to detect this disease, deep learning algorithms and machine vision are widely used by computer scientists. Convolutional neural networks (CNN), DenseNet121, Resnet50, and VGG16 were used in this study for the detection of Covid-19 in X-ray images. In the current study, 1341 chest radiographs from the COVID-19 dataset were used to detect COVID-19 including infected and Healthy classes using a modified pre-trained CNN (train and test accuracy of 99.75% and 99.63%, respectively). The DENSENET121 model has a training accuracy of 43.89% and a test accuracy of 57.89%, respectively. The train and test accuracy of ResNet-50 are, respectively, 89.43% and 90%. Additionally, the CNN model has test and train accuracy of 98.13% and 96.73%, respectively. The suggested model has COVID-19 detection accuracy that is at least 1% higher than all other models.