A Shallow–Deep Feature Fusion Method for Pedestrian Detection

In this paper, a shallow–deep feature fusion (SDFF) method is developed for pedestrian detection. Firstly, we propose a shallow feature-based method under the ACF framework of pedestrian detection. More precisely, improved Haar-like templates with Local FDA learning are used to filter the channel maps of ACF such that these Haar-like features are able to improve the discriminative power and therefore enhance the detection performance. The proposed shallow feature is also referred to as weighted subset-haar-like feature. It is efficient in pedestrian detection with a high recall rate and precise localization. Secondly, the proposed shallow feature-based detection method operates as a region proposal. A classifier equipped with ResNet is then used to refine the region proposals to judge whether each region contains a pedestrian or not. The extensive experiments evaluated on INRIA, Caltech, and TUD-Brussel datasets show that SDFF is an effective and efficient method for pedestrian detection.

Exploring highly reliable substructures in auto-reconstructions of a neuron

The digital reconstruction of a neuron is the most direct and effective way to investigate its morphology. Many automatic neuron tracing methods have been proposed, but without manual check it is difficult to know whether a reconstruction or which substructure in a reconstruction is accurate. For a neuron’s reconstructions generated by multiple automatic tracing methods with different principles or models, their common substructures are highly reliable and named individual motifs. In this work, we propose a Vaa3D-based method called Lamotif to explore individual motifs in automatic reconstructions of a neuron. Lamotif utilizes the local alignment algorithm in BlastNeuron to extract local alignment pairs between a specified objective reconstruction and multiple reference reconstructions, and combines these pairs to generate individual motifs on the objective reconstruction. The proposed Lamotif is evaluated on reconstructions of 163 multiple species neurons, which are generated by four state-of-the-art tracing methods. Experimental results show that individual motifs are almost on corresponding gold standard reconstructions and have much higher precision rate than objective reconstructions themselves. Furthermore, an objective reconstruction is mostly quite accurate if its individual motifs have high recall rate. Individual motifs contain common geometry substructures in multiple reconstructions, and can be used to select some accurate substructures from a reconstruction or some accurate reconstructions from automatic reconstruction dataset of different neurons.

Comparison of Mammography and Ultrasonography for Early Detection of Breast Cancer

Aim: This systematic review is specifically aimed to compare mammography and ultrasonography in early detection of breast cancer. For this systematic review, major purpose is to compare both screening methods and also analyze the performance of supplemental ultrasonography for early detection of breast cancer. Methodology: For this systematic review, total 23 studies are included which follow the Preferred Reporting Items for Systematic Reviews and Meta-Analysis (PRISMA) guidelines. Electronic articles from year 2007 to from year 2017 on PUB Med, online Willey library, and Science Direct site were searched by using keywords related to sonographic and mammography imaging for breast cancer. Results: Out of 23 studies, 12 studies are conducted on women with dense breasts. Twenty studies performed their imaging with hand held ultrasound (HHUS). Out of twenty-three studies, sixteen studies followed BI-RADS procedures. In eleven studies that used joint methods, it was observed that mammography (MAM) has 65% whereas ultrasound (US) has 68% efficiency for early detection of breast cancer. 88% area under a cover (AUCs) among MAM and 98% among US imaging was observed. No major difference was found in sensitivity and specificity of both techniques. Conclusion: Study concludes that Ultrasound is more efficient to diagnose factors suggestive of breast cancer that cannot be detected on mammography. It also has the potential to evaluate cancer among dense breast women but unfortunately in some cases, it may cause a high recall rate. Keywords: Breast, Cancer, Mammography, Ultrasonography, Screening.

Exploring high reliable substructures in auto-reconstructions of a neuron

The digital reconstruction of a neuron is the most direct and effective way to investigate its morphology. Many automatic neuron tracing methods have been proposed, but without manual check it is difficult to know whether a reconstruction or which substructure in a reconstruction is accurate. For a neuron’s reconstructions generated by multiple automatic tracing methods with different principles or models, their common substructures are highly reliable and named individual motifs. In this work, we propose a Vaa3D based method called Lamotif to explore individual motifs in automatic reconstructions of a neuron. Lamotif utilizes the local alignment algorithm in BlastNeuron to extract local alignment pairs between a specified objective reconstruction and multiple reference reconstructions, and combines these pairs to generate individual motifs on the objective reconstruction. The proposed Lamotif is evaluated on reconstructions of 163 multiple species neurons, which are generated by four state-of-the-art tracing methods. Experimental results show that individual motifs are almost on corresponding gold standard reconstructions and have much higher precision rate than objective reconstructions themselves. Furthermore, an objective reconstruction is mostly quite accurate if its individual motifs have high recall rate. Individual motifs contain common geometry substructures in multiple reconstructions, and can be used to select some accurate substructures from a reconstruction or some accurate reconstructions from automatic reconstruction dataset of different neurons.

Malicious URL Detection Based on Improved Multilayer Recurrent Convolutional Neural Network Model

The traditional malicious uniform resource locator (URL) detection method excessively relies on the matching rules formulated by the network security personnel, which is hard to fully express the text information of the URL. Thus, an improved multilayer recurrent convolutional neural network model based on the YOLO algorithm is proposed to detect malicious URL in this paper. First, single characters are mapped to dense vectors using word embedding, and the dense vectors are participated in the training process of the whole model according to the structural characteristics of the URL in the method. Then, the CSPDarknet neural network model based on the improved YOLO algorithm is proposed to extract features of the URL. Finally, the extracted features are used to evaluate malicious URL by the bidirectional LSTM recurrent neural network algorithm. In order to verify the validity of the algorithm, a total of 200,000 URLs are collected, including 100,000 normal URLs labeled “good” and 100,000 malicious URLs labeled “bad”. The experimental results show that the method detects malicious URLs more quickly and effectively and has high accuracy, high recall rate, and high accuracy compared with Text-RCNN, BRNN, and other models.

PurExt: Automated Extraction of the Purpose-Aware Rule from the Natural Language Privacy Policy in IoT

The extensive data collection performed by the Internet of Things (IoT) devices can put users at risk of data leakage. Consequently, IoT vendors are legally obliged to provide privacy policies to declare the scope and purpose of the data collection. However, complex and lengthy privacy policies are unfriendly to users, and the lack of a machine-readable format makes it difficult to check policy compliance automatically. To solve these problems, we first put forward a purpose-aware rule to formalize the purpose-driven data collection or use statement. Then, a novel approach to identify the rule from natural language privacy policies is proposed. To address the issue of diversity of purpose expression, we present the concepts of explicit and implicit purpose, which enable using the syntactic and semantic analyses to extract purposes in different sentences. Finally, the domain adaption method is applied to the semantic role labeling (SRL) model to improve the efficiency of purpose extraction. The experiments that are conducted on the manually annotated dataset demonstrate that this approach can extract purpose-aware rules from the privacy policies with a high recall rate of 91%. The implicit purpose extraction of the adapted model significantly improves the F1-score by 11%.

Evaluation of the Microba Community Profiler for Taxonomic Profiling of Metagenomic Datasets From the Human Gut Microbiome

A fundamental goal of microbial ecology is to accurately determine the species composition in a given microbial ecosystem. In the context of the human microbiome, this is important for establishing links between microbial species and disease states. Here we benchmark the Microba Community Profiler (MCP) against other metagenomic classifiers using 140 moderate to complex in silico microbial communities and a standardized reference genome database. MCP generated accurate relative abundance estimates and made substantially fewer false positive predictions than other classifiers while retaining a high recall rate. We further demonstrated that the accuracy of species classification was substantially increased using the Microba Genome Database, which is more comprehensive than reference datasets used by other classifiers and illustrates the importance of including genomes of uncultured taxa in reference databases. Consequently, MCP classifies appreciably more reads than other classifiers when using their recommended reference databases. These results establish MCP as best-in-class with the ability to produce comprehensive and accurate species profiles of human gastrointestinal samples.

A Data Mining Method for Students' Behavior Understanding

To model students' behavior and describe their behavior characteristics accurately and comprehensively, a framework for predicting students' learning performance based on behavioral model is proposed, which extracts features from multiple perspectives to describe behaviors more comprehensively, including statistical features and association features. In addition, a multi-task model is designed for fine-grained prediction of students' learning performance in the curriculum. A framework for predicting mastery based on online learning behavior is also put forward. Additional context information is added to the collaborative filtering algorithm, including student-knowledge-point mastery and class-knowledge-point, and students' mastery is predicted according to the learning path excavated. Considering the time-varying of mastery, the approximate curve of students' mastery of knowledge points is fitted according to the Ebinhaus forgetting curve. The experiments show that the proposed framework has a high recall rate for the prediction of learning performance, and also shows a certain practicability for early warning. Further, based on the model, the correlation between student behavior patterns and learning performance is discussed. The addition of additional information has improved the prediction efficiency, especially the operational efficiency. At the same time, the proposed framework can not only dynamically assess students' master of knowledge, but also facilitate the system to review feedback or adjust the learning order, and provide personalized learning services.

Intelligent Driver Warning System using Deep Learning-based Facial Expression Recognition

Driver’s inattention is one of the major factors and reasons in occurrence of many road accidents and unforeseen crashes. Hence it is crucial to develop an automatic driver warning system that can send timely warning signals to the drivers. This issue involves determining the driver’s mental state that is ultimately based on the driver’s facial expressions. Automated facial emotion recognition is a recent development in the image processing domain and is the need of the hour in applications like driver warning systems. The existing methods are capable of recognizing facial emotions even when provided with a noisy signal or imperfect data, but ultimately it lacks accuracy. It is also ineffective in dealing with spontaneous emotions, and recognition. The proposed approach develops a driver warning system that extracts the facial expressions based on a novel efficient Local Octal Pattern (LOP) and effectively recognizes the facial expressions based on Deep Neural Networks, Convolutional Neural Networks (CNN). The LOP feature map serves as an input to CNN and guides in the selection of CNN learning data thereby improving and further enhancing the understanding and learning of CNN. It also has an ability to recognize both natural and spontaneous emotions, as well as image and video can be considered as an input.The experimental results consideringYawDD dataset indicates that the proposed system has been efficiently evaluated by considering the with metrics such as Precision, Recall and F-Score and thereby it is observed and inferred that the proposed system obtained a high recall rate of 96.09% in comparison with the other state-of-the-art methods

SOFIA: A Novel Automated Breast Ultrasound System Used on Patients in the Prone Position: A Pilot Study on Lesion Detection in Comparison to Handheld Grayscale Ultrasound

Introduction Most of the currently available automated breast ultrasound systems require patients to be in the supine position. Previous data, however, show a high recall rate with this method due to artifacts. The novel automated breast ultrasound scanner SOFIA scans the breast with the patient in a prone position, resulting in even compression of breast tissue. We present our initial results with this examination method. Material and Methods 63 patients were analyzed using a handheld B-mode ultrasound. In cases of BI-RADS 1, 2 or 5, a SOFIA scan was performed. Sensitivity, specificity and accuracy were calculated. Interobserver agreement was evaluated using Cohenʼs kappa. The duration of the scan was measured for both methods. Results No BI-RADS 5 lesion was missed with SOFIA. The SOFIA had an additional recall rate of 16.67% compared to B-mode ultrasound. The sensitivity, specificity and accuracy of SOFIA was 100, 83.33 and 88.89%, respectively. Cohenʼs kappa showed substantial agreement (κ = 0.769) between examiner 1 (B-mode) and examiner 2 (SOFIA). The mean scan duration for the B-mode system and the SOFIA system was 24.21 minutes and 12.94 minutes, respectively. In four cases, D-cup breasts were not scanned in their entirety. Conclusion No cancer was missed when SOFIA was used in this preselected study population. The scanning time was approximately half of that required for B-mode ultrasound. The additional unnecessary recall rate was 16.67%. Larger D cup-size breasts were difficult to position and resulted in an incomplete image in four cases.