Search results for: feature detection

Authors: Maha Jazouli

Abstract:

Suicide is one of the most important causes of death in the prison environment, both in Canada and internationally. Rates of attempts of suicide and self-harm have been on the rise in recent years, with hangings being the most frequent method resorted to. The objective of this article is to propose a method to automatically detect in real time suicidal behaviors. We present a gesture recognition system that consists of three modules: model-based movement tracking, feature extraction, and gesture recognition using machine learning algorithms (MLA). Our proposed system gives us satisfactory results. This smart video surveillance system can help assist staff responsible for the safety and health of inmates by alerting them when suicidal behavior is detected, which helps reduce mortality rates and save lives.

Keywords: suicide detection, Kinect azure, RGB-D camera, SVM, machine learning, gesture recognition

Procedia PDF Downloads 148

Authors: Robert Caulk

Abstract:

A generalized framework for adaptive machine learning deployments in algorithmic trading is introduced, tested, and released as open-source code. The presented software aims to test the hypothesis that recent data contains enough information to form a probabilistically favorable short-term price prediction. Further, the framework contains various adaptive machine learning techniques that are geared toward generating profit during strong trends and minimizing losses during trend changes. Results demonstrate that this adaptive machine learning approach is capable of capturing trends and generating profit. The presentation also discusses the importance of defining the parameter space associated with the dynamic training data-set and using the parameter space to identify and remove outliers from prediction data points. Meanwhile, the generalized architecture enables common users to exploit the powerful machinery while focusing on high-level feature engineering and model testing. The presentation also highlights common strengths and weaknesses associated with the presented technique and presents a broad range of well-tested starting points for feature set construction, target setting, and statistical methods for enforcing risk management and maintaining probabilistically favorable entry and exit points. The presentation also describes the end-to-end data processing tools associated with FreqAI, including automatic data fetching, data aggregation, feature engineering, safe and robust data pre-processing, outlier detection, custom machine learning and statistical tools, data post-processing, and adaptive training backtest emulation, and deployment of adaptive training in live environments. Finally, the generalized user interface is also discussed in the presentation. Feature engineering is simplified so that users can seed their feature sets with common indicator libraries (e.g. TA-lib, pandas-ta). The user also feeds data expansion parameters to fill out a large feature set for the model, which can contain as many as 10,000+ features. The presentation describes the various object-oriented programming techniques employed to make FreqAI agnostic to third-party libraries and external data sources. In other words, the back-end is constructed in such a way that users can leverage a broad range of common regression libraries (Catboost, LightGBM, Sklearn, etc) as well as common Neural Network libraries (TensorFlow, PyTorch) without worrying about the logistical complexities associated with data handling and API interactions. The presentation finishes by drawing conclusions about the most important parameters associated with a live deployment of the adaptive learning framework and provides the road map for future development in FreqAI.

Keywords: machine learning, market trend detection, open-source, adaptive learning, parameter space exploration

Procedia PDF Downloads 61

Authors: Samane Sharifi Monfared, Lavdie Rada

Abstract:

Nowadays, autonomous vehicles are developing rapidly toward facilitating human car driving. One of the main issues is road lane detection for a suitable guidance direction and car accident prevention. This paper aims to improve and optimize road line detection based on a combination of camera calibration, the Hough transform, and Canny edge detection. The video processing is implemented using the Open CV library with the novelty of having a scale able region masking. The aim of the study is to introduce automatic road lane detection techniques with the user’s minimum manual intervention.

Keywords: hough transform, canny edge detection, optimisation, scaleable masking, camera calibration, improving the quality of image, image processing, video processing

Procedia PDF Downloads 63

Authors: Adrià Arbués-Sangüesa, Coloma Ballester, Gloria Haro

Abstract:

Tracking sports players is a widely challenging scenario, specially in single-feed videos recorded in tight courts, where cluttering and occlusions cannot be avoided. This paper presents an analysis of several geometric and semantic visual features to detect and track basketball players. An ablation study is carried out and then used to remark that a robust tracker can be built with Deep Learning features, without the need of extracting contextual ones, such as proximity or color similarity, nor applying camera stabilization techniques. The presented tracker consists of: (1) a detection step, which uses a pretrained deep learning model to estimate the players pose, followed by (2) a tracking step, which leverages pose and semantic information from the output of a convolutional layer in a VGG network. Its performance is analyzed in terms of MOTA over a basketball dataset with more than 10k instances.

Keywords: basketball, deep learning, feature extraction, single-camera, tracking

Procedia PDF Downloads 108

Authors: Xiaoheng Tan, Xianfang Li, Tan Guo, Yuchuan Liu, Zhijun Yang, Hongye Li, Kai Fu, Yufang Wu, Heling Gong

Abstract:

The image data collected in reality often have high dimensions, and it contains noise and redundant information. Therefore, it is necessary to extract the compact feature expression of the original perceived image. In this process, effective use of prior knowledge such as data structure distribution and sample label is the key to enhance image feature discrimination and robustness. Based on the above considerations, this paper proposes a local preserving discriminant feature learning model based on graph optimization. The model has the following characteristics: (1) Locality preserving constraint can effectively excavate and preserve the local structural relationship between data. (2) The flexibility of graph learning can be improved by constructing a new local geometric structure graph using label information and the nearest neighbor threshold. (3) The L₂,₁ norm is used to redefine LDA, and the diagonal matrix is introduced as the scale factor of LDA, and the samples are selected, which improves the robustness of feature learning. The validity and robustness of the proposed algorithm are verified by experiments in two public image datasets.

Keywords: feature extraction, graph optimization local preserving projection, linear discriminant analysis, L₂, ₁ norm

Procedia PDF Downloads 117

Authors: Avatharam Ganivada, Krishna Shah

Abstract:

The Mask R-CNN is recently introduced by the team of Facebook AI Research (FAIR), which is mainly concerned with instance segmentation in images. Here, the Mask R-CNN is based on ResNet and feature pyramid network (FPN), where a single dropout method is employed. This paper provides a modified Mask R-CNN by adding multiple dropout methods into the Mask R-CNN. The proposed model has also utilized the concepts of Resnet and FPN to extract stage-wise network feature maps, wherein a top-down network path having lateral connections is used to obtain semantically strong features. The proposed model produces three outputs for each object in the image: class label, bounding box coordinates, and object mask. The performance of the proposed network is evaluated in the segmentation of every instance in images using COCO and cityscape datasets. The proposed model achieves better performance than the state-of-the-networks for the datasets.

Keywords: instance segmentation, object detection, convolutional neural networks, deep learning, computer vision

Procedia PDF Downloads 41

Authors: Xiao Chen, Xiaoying Kong, Min Xu

Abstract:

This paper presents a road vehicle detection approach for the intelligent transportation system. This approach mainly uses low-cost magnetic sensor and associated data collection system to collect magnetic signals. This system can measure the magnetic field changing, and it also can detect and count vehicles. We extend Mel Frequency Cepstral Coefficients to analyze vehicle magnetic signals. Vehicle type features are extracted using representation of cepstrum, frame energy, and gap cepstrum of magnetic signals. We design a 2-dimensional map algorithm using Vector Quantization to classify vehicle magnetic features to four typical types of vehicles in Australian suburbs: sedan, VAN, truck, and bus. Experiments results show that our approach achieves a high level of accuracy for vehicle detection and classification.

Keywords: vehicle classification, signal processing, road traffic model, magnetic sensing

Procedia PDF Downloads 291

Authors: Tudor Barbu

Abstract:

An automatic moment-based texture segmentation approach is proposed in this paper. First, we describe the related work in this computer vision domain. Our texture feature extraction, the first part of the texture recognition process, produces a set of moment-based feature vectors. For each image pixel, a texture feature vector is computed as a sequence of area moments. Second, an automatic pixel classification approach is proposed. The feature vectors are clustered using some unsupervised classification algorithm, the optimal number of clusters being determined using a measure based on validation indexes. From the resulted pixel classes one determines easily the desired texture regions of the image.

Keywords: image segmentation, moment-based, texture analysis, automatic classification, validation indexes

Procedia PDF Downloads 383

Authors: Jigg L. Pelayo, Ricardo G. Villar, Einstine M. Opiso

Abstract:

The advances of aerial laser scanning in the Philippines has open-up entire fields of research in remote sensing and machine vision aspire to provide accurate timely information for the government and the public. Rapid mapping of polygonal roads and roof boundaries is one of its utilization offering application to disaster risk reduction, mitigation and development. The study uses low density LiDAR data and high resolution aerial imagery through object-oriented approach considering the theoretical concept of data analysis subjected to machine learning algorithm in minimizing the constraints of feature extraction. Since separating one class from another in distinct regions of a multi-dimensional feature-space, non-trivial computing for fitting distribution were implemented to formulate the learned ideal hyperplane. Generating customized hybrid feature which were then used in improving the classifier findings. Supplemental algorithms for filtering and reshaping object features are develop in the rule set for enhancing the final product. Several advantages in terms of simplicity, applicability, and process transferability is noticeable in the methodology. The algorithm was tested in the different random locations of Misamis Oriental province in the Philippines demonstrating robust performance in the overall accuracy with greater than 89% and potential to semi-automation. The extracted results will become a vital requirement for decision makers, urban planners and even the commercial sector in various assessment processes.

Keywords: feature extraction, machine learning, OBIA, remote sensing

Procedia PDF Downloads 333

Authors: Mst Shapna Akter, Hossain Shahriar

Abstract:

One of the most important challenges in the field of software code audit is the presence of vulnerabilities in software source code. Every year, more and more software flaws are found, either internally in proprietary code or revealed publicly. These flaws are highly likely exploited and lead to system compromise, data leakage, or denial of service. C and C++ open-source code are now available in order to create a largescale, machine-learning system for function-level vulnerability identification. We assembled a sizable dataset of millions of opensource functions that point to potential exploits. We developed an efficient and scalable vulnerability detection method based on deep neural network models that learn features extracted from the source codes. The source code is first converted into a minimal intermediate representation to remove the pointless components and shorten the dependency. Moreover, we keep the semantic and syntactic information using state-of-the-art word embedding algorithms such as glove and fastText. The embedded vectors are subsequently fed into deep learning networks such as LSTM, BilSTM, LSTM-Autoencoder, word2vec, BERT, and GPT-2 to classify the possible vulnerabilities. Furthermore, we proposed a neural network model which can overcome issues associated with traditional neural networks. Evaluation metrics such as f1 score, precision, recall, accuracy, and total execution time have been used to measure the performance. We made a comparative analysis between results derived from features containing a minimal text representation and semantic and syntactic information. We found that all of the deep learning models provide comparatively higher accuracy when we use semantic and syntactic information as the features but require higher execution time as the word embedding the algorithm puts on a bit of complexity to the overall system.

Keywords: cyber security, vulnerability detection, neural networks, feature extraction

Procedia PDF Downloads 44

Authors: Ugochukwu Ejike Akpudo, Jank-Wook Hur

Abstract:

The increasing demand for improved productivity, maintainability, and reliability has prompted rapidly increasing research studies on the emerging condition-based maintenance concept- Prognostics and health management (PHM). Varieties of fuel pumps serve critical functions in several hydraulic systems; hence, their failure can have daunting effects on productivity, safety, etc. The need for condition monitoring and assessment of these pumps cannot be overemphasized, and this has led to the uproar in research studies on standard feature extraction techniques for optimized condition assessment of fuel pumps. By extracting time-based, frequency-based and the more robust time-frequency based features from these vibrational signals, a more comprehensive feature assessment (and selection) can be achieved for a more accurate and reliable condition assessment of these pumps. With the aid of emerging deep classification and regression algorithms like the locally linear embedding (LLE), we propose a method for comprehensive condition assessment of electromagnetic fuel pumps (EMFPs). Results show that the LLE as a comprehensive feature extraction technique yields better feature fusion/dimensionality reduction results for condition assessment of EMFPs against the use of single features. Also, unlike other feature fusion techniques, its capabilities as a fault classification technique were explored, and the results show an acceptable accuracy level using standard performance metrics for evaluation.

Keywords: electromagnetic fuel pumps, comprehensive feature extraction, condition assessment, locally linear embedding, feature fusion

Procedia PDF Downloads 87

Authors: Ainagul Yermekova, Liudmila Goncharenko, Ali Baghirzade, Sergey Sybachin

Abstract:

In this article, we address the problem of Kazakhstan's transition to the Latin alphabet. The transition process started in 2017 and is scheduled to be completed in 2025. In connection with these events, the problem of recognizing the characters of the new alphabet is raised. Well-known character recognition programs such as ABBYY FineReader, FormReader, MyScript Stylus did not recognize specific Kazakh letters that were used in Cyrillic. The author tries to give an assessment of the well-known method of character recognition that could be in demand as part of the country's transition to the Latin alphabet. Three methods of character recognition: template, structured, and feature-based, are considered through the algorithms of operation. At the end of the article, a general conclusion is made about the possibility of applying a certain method to a particular recognition process: for example, in the process of population census, recognition of typographic text in Latin, or recognition of photos of car numbers, store signs, etc.

Keywords: text detection, template method, recognition algorithm, structured method, feature method

Procedia PDF Downloads 154

Authors: Xiaoqing Wang, Junfeng Wang, Xiaolan Zhu

Abstract:

Android operating system has been recognized by most application developers because of its good open-source and compatibility, which enriches the categories of applications greatly. However, it has become the target of malware attackers due to the lack of strict security supervision mechanisms, which leads to the rapid growth of malware, thus bringing serious safety hazards to users. Therefore, it is critical to detect Android malware effectively. Generally, the permissions declared in the AndroidManifest.xml can reflect the function and behavior of the application to a large extent. Since current Android system has not any restrictions to the number of permissions that an application can request, developers tend to apply more than actually needed permissions in order to ensure the successful running of the application, which results in the abuse of permissions. However, some traditional detection methods only consider the requested permissions and ignore whether it is actually used, which leads to incorrect identification of some malwares. Therefore, a machine learning detection method based on the actually used permissions combination and API calls was put forward in this paper. Meanwhile, several experiments are conducted to evaluate our methodology. The result shows that it can detect unknown malware effectively with higher true positive rate and accuracy while maintaining a low false positive rate. Consequently, the AdaboostM1 (J48) classification algorithm based on information gain feature selection algorithm has the best detection result, which can achieve an accuracy of 99.8%, a true positive rate of 99.6% and a lowest false positive rate of 0.

Keywords: android, API Calls, machine learning, permissions combination

Procedia PDF Downloads 304

Authors: Rohit Gupta, Ravinder Agarwal

Abstract:

Surface electromyographic (sEMG) signal has the potential to identify the human activities and intention. This potential is further exploited to control the artificial limbs using the sEMG signal from residual limbs of amputees. The paper deals with the development of multichannel cost efficient sEMG signal interface for research application, along with evaluation of proposed class dependent statistical approach of the feature selection method. The sEMG signal acquisition interface was developed using ADS1298 of Texas Instruments, which is a front-end interface integrated circuit for ECG application. Further, the sEMG signal is recorded from two lower limb muscles for three locomotions namely: Plane Walk (PW), Stair Ascending (SA), Stair Descending (SD). A class dependent statistical approach is proposed for feature selection and also its performance is compared with 12 preexisting feature vectors. To make the study more extensive, performance of five different types of classifiers are compared. The outcome of the current piece of work proves the suitability of the proposed feature selection algorithm for locomotion recognition, as compared to other existing feature vectors. The SVM Classifier is found as the outperformed classifier among compared classifiers with an average recognition accuracy of 97.40%. Feature vector selection emerges as the most dominant factor affecting the classification performance as it holds 51.51% of the total variance in classification accuracy. The results demonstrate the potentials of the developed sEMG signal acquisition interface along with the proposed feature selection algorithm.

Keywords: classifiers, feature selection, locomotion, sEMG

Procedia PDF Downloads 262

Authors: M. Kowalski, M. Kastek, M. Szustakowski

Abstract:

Multispectral screening systems are becoming more popular because of their very interesting properties and applications. One of the most significant applications of multispectral screening systems is prevention of terrorist attacks. There are many kinds of threats and many methods of detection. Visual detection of objects hidden under clothing of a person is one of the most challenging problems of threats detection. There are various solutions of the problem; however, the most effective utilize multispectral surveillance imagers. The development of imaging devices and exploration of new spectral bands is a chance to introduce new equipment for assuring public safety. We investigate the possibility of long lasting detection of potentially dangerous objects covered with various types of clothing. In the article we present the results of comparative studies of passive imaging in three spectrums – visible, infrared and terahertz

Keywords: terahertz, infrared, object detection, screening camera, image processing

Procedia PDF Downloads 321

Authors: Yao-Hong Tsai

Abstract:

This paper presents an efficient fusion algorithm for iris images to generate stable feature for recognition in unconstrained environment. Recently, iris recognition systems are focused on real scenarios in our daily life without the subject’s cooperation. Under large variation in the environment, the objective of this paper is to combine information from multiple images of the same iris. The result of image fusion is a new image which is more stable for further iris recognition than each original noise iris image. A wavelet-based approach for multi-resolution image fusion is applied in the fusion process. The detection of the iris image is based on Adaboost algorithm and then local binary pattern (LBP) histogram is then applied to texture classification with the weighting scheme. Experiment showed that the generated features from the proposed fusion algorithm can improve the performance for verification system through iris recognition.

Keywords: image fusion, iris recognition, local binary pattern, wavelet

Procedia PDF Downloads 342

Authors: Zouhair Haddi, Stephane Delliaux, Jean-Francois Pons, Ismail Kechaf, Jean-Claude De Haro, Mustapha Ouladsine

Abstract:

Atrial fibrillation (AF) has been considered as the most common cardiac arrhythmia, and a major public health burden associated with significant morbidity and mortality. Nowadays, telemedical approaches targeting cardiac outpatients situate AF among the most challenged medical issues. The automatic, early, and fast AF detection is still a major concern for the healthcare professional. Several algorithms based on univariate analysis have been developed to detect atrial fibrillation. However, the published results do not show satisfactory classification accuracy. This work was aimed at resolving this shortcoming by proposing multivariate data analysis methods for automatic AF detection. Four publicly-accessible sets of clinical data (AF Termination Challenge Database, MIT-BIH AF, Normal Sinus Rhythm RR Interval Database, and MIT-BIH Normal Sinus Rhythm Databases) were used for assessment. All time series were segmented in 1 min RR intervals window and then four specific features were calculated. Two pattern recognition methods, i.e., Principal Component Analysis (PCA) and Learning Vector Quantization (LVQ) neural network were used to develop classification models. PCA, as a feature reduction method, was employed to find important features to discriminate between AF and Normal Sinus Rhythm. Despite its very simple structure, the results show that the LVQ model performs better on the analyzed databases than do existing algorithms, with high sensitivity and specificity (99.19% and 99.39%, respectively). The proposed AF detection holds several interesting properties, and can be implemented with just a few arithmetical operations which make it a suitable choice for telecare applications.

Keywords: atrial fibrillation, multivariate data analysis, automatic detection, telemedicine

Procedia PDF Downloads 237

Authors: Seyed-Yaser Nabavi-Chashmi, Davood Asadi, Karim Ahmadi, Eren Demir

Abstract:

The landing phase of a UAV is very critical as there are many uncertainties in this phase, which can easily entail a hard landing or even a crash. In this paper, the estimation of relative distance and velocity to the ground, as one of the most important processes during the landing phase, is studied. Using accurate measurement sensors as an alternative approach can be very expensive for sensors like LIDAR, or with a limited operational range, for sensors like ultrasonic sensors. Additionally, absolute positioning systems like GPS or IMU cannot provide distance to the ground independently. The focus of this paper is to determine whether we can measure the relative distance and velocity of UAV and ground in the landing phase using just low-resolution images taken by a monocular camera. The Lucas-Konda feature detection technique is employed to extract the most suitable feature in a series of images taken during the UAV landing. Two different approaches based on Extended Kalman Filters (EKF) have been proposed, and their performance in estimation of the relative distance and velocity are compared. The first approach uses the kinematics of the UAV as the process and the calculated optical flow as the measurement; On the other hand, the second approach uses the feature’s projection on the camera plane (pixel position) as the measurement while employing both the kinematics of the UAV and the dynamics of variation of projected point as the process to estimate both relative distance and relative velocity. To verify the results, a sequence of low-quality images taken by a camera that is moving on a specifically developed testbed has been used to compare the performance of the proposed algorithm. The case studies show that the quality of images results in considerable noise, which reduces the performance of the first approach. On the other hand, using the projected feature position is much less sensitive to the noise and estimates the distance and velocity with relatively high accuracy. This approach also can be used to predict the future projected feature position, which can drastically decrease the computational workload, as an important criterion for real-time applications.

Keywords: altitude estimation, drone, image processing, trajectory planning

Procedia PDF Downloads 80

Authors: Tesnim Charrad, Kaouther Nouira, Ahmed Ferchichi

Abstract:

In order to reduce the number of deaths due to heart problems, we propose the use of Hierarchical Temporal Memory Algorithm (HTM) which is a real time anomaly detection algorithm. HTM is a cortical learning algorithm based on neocortex used for anomaly detection. In other words, it is based on a conceptual theory of how the human brain can work. It is powerful in predicting unusual patterns, anomaly detection and classification. In this paper, HTM have been implemented and tested on ECG datasets in order to detect cardiac anomalies. Experiments showed good performance in terms of specificity, sensitivity and execution time.

Keywords: cardiac anomalies, ECG, HTM, real time anomaly detection

Procedia PDF Downloads 185

Authors: Yousef Farhaoui

Abstract:

An IDS is a tool which is used to improve the level of security.In this paper we present different architectures of IDS. We will also discuss measures that define the effectiveness of IDS and the very recent works of standardization and homogenization of IDS. At the end, we propose a new model of IDS called BiIDS (IDS Based on the two principles of detection).

Keywords: intrusion detection, architectures, characteristic, tools, security

Procedia PDF Downloads 430

Authors: Wafa' Slaibi Alsharafat

Abstract:

Cloud computing is a modern approach in network environment. According to increased number of network users and online systems, there is a need to help these systems to be away from unauthorized resource access and detect any attempts for privacy contravention. For that purpose, Intrusion Detection System is an effective security mechanism to detect any attempts of attacks for cloud resources and their information. In this paper, Cloud Intrusion Detection System has been proposed in term of reducing or eliminating any attacks. This model concerns about achieving high detection rate after conducting a set of experiments using benchmarks dataset called KDD'99.

Keywords: IDS, cloud computing, anticipating classifier system, intrusion detection

Procedia PDF Downloads 445

Authors: Heramb Somthankar, Anujoy Chakraborty

Abstract:

This paper presents a classification study on yield prediction in semiconductor production using machine learning approaches. A complicated semiconductor production process is generally monitored continuously by signals acquired from sensors and measurement sites. A monitoring system contains a variety of signals, all of which contain useful information, irrelevant information, and noise. In the case of each signal being considered a feature, "Feature Selection" is used to find the most relevant signals. The open-source UCI SECOM Dataset provides 1567 such samples, out of which 104 fail in quality assurance. Feature extraction and selection are performed on the dataset, and useful signals were considered for further study. Afterward, common machine learning algorithms were employed to predict whether the signal yields pass or fail. The most relevant algorithm is selected for prediction based on the accuracy and loss of the ML model.

Keywords: deep learning, feature extraction, feature selection, machine learning classification algorithms, semiconductor production monitoring, signal processing, time-series analysis

Procedia PDF Downloads 68

Authors: Tatsuya Takino, Izuru Nomura, Yuji Kageyama, Shin Nagata, Hiroyuki Kamata

Abstract:

We propose a method of detecting the craters from the image of the lunar surface. This proposal assumes that it is applied to SLIM (Smart Lander for Investigating Moon) working group aiming at the pinpoint landing on the lunar surface and investigating scientific research. It is difficult to equip and use high-performance computers for the small space probe. So, it is necessary to use a small computer with an exclusive hardware such as FPGA. We have studied the crater detection using principal component analysis (PCA), In this paper, We implement detection algorithm into the FPGA, and the detection is performed on the data that was captured from the CMOS camera.

Keywords: crater detection, PCA, FPGA, image processing

Procedia PDF Downloads 514

Authors: Armin Nabaei, M. Omair Ahmad, M. N. S. Swamy

Abstract:

Many facial expression and emotion recognition methods in the traditional approaches of using LDA, PCA, and EBGM have been proposed. In recent years deep learning models have provided a unique platform addressing by automatically extracting the features for the detection of facial expression and emotions. However, deep networks require large training datasets to extract automatic features effectively. In this work, we propose an efficient emotion detection algorithm using face images when only small datasets are available for training. We design a deep network whose feature extraction capability is enhanced by utilizing several parallel modules between the input and output of the network, each focusing on the extraction of different types of coarse features with fined grained details to break the symmetry of produced information. In fact, we leverage long range dependencies, which is one of the main drawback of CNNs. We develop this work by introducing a Dynamic Soft-Margin SoftMax.The conventional SoftMax suffers from reaching to gold labels very soon, which take the model to over-fitting. Because it’s not able to determine adequately discriminant feature vectors for some variant class labels. We reduced the risk of over-fitting by using a dynamic shape of input tensor instead of static in SoftMax layer with specifying a desired Soft- Margin. In fact, it acts as a controller to how hard the model should work to push dissimilar embedding vectors apart. For the proposed Categorical Loss, by the objective of compacting the same class labels and separating different class labels in the normalized log domain.We select penalty for those predictions with high divergence from ground-truth labels.So, we shorten correct feature vectors and enlarge false prediction tensors, it means we assign more weights for those classes with conjunction to each other (namely, “hard labels to learn”). By doing this work, we constrain the model to generate more discriminate feature vectors for variant class labels. Finally, for the proposed optimizer, our focus is on solving weak convergence of Adam optimizer for a non-convex problem. Our noteworthy optimizer is working by an alternative updating gradient procedure with an exponential weighted moving average function for faster convergence and exploiting a weight decay method to help drastically reducing the learning rate near optima to reach the dominant local minimum. We demonstrate the superiority of our proposed work by surpassing the first rank of three widely used Facial Expression Recognition datasets with 93.30% on FER-2013, and 16% improvement compare to the first rank after 10 years, reaching to 90.73% on RAF-DB, and 100% k-fold average accuracy for CK+ dataset, and shown to provide a top performance to that provided by other networks, which require much larger training datasets.

Keywords: computer vision, facial expression recognition, machine learning, algorithms, depp learning, neural networks

Procedia PDF Downloads 48

Authors: Yuanchao Sang, Zhihao Gong, Longsheng Chen, Long Chen

Abstract:

Clothing fashion represents a human’s aesthetic appreciation towards everyday outfits and appetite for fashion, and it reflects the development of status in society, humanity, and economics. However, modelling fashion by machine is extremely challenging because fashion is too abstract to be efficiently described by machines. Even human beings can hardly reach a consensus about fashion. In this paper, we are dedicated to answering a fundamental fashion-related problem: what image feature best describes clothing fashion? To address this issue, we have designed and evaluated various image features, ranging from traditional low-level hand-crafted features to mid-level style awareness features to various current popular deep neural network-based features, which have shown state-of-the-art performance in various vision tasks. In summary, we tested the following 9 feature representations: color, texture, shape, style, convolutional neural networks (CNNs), CNNs with distance metric learning (CNNs&DML), AutoEncoder, CNNs with multiple layer combination (CNNs&MLC) and CNNs with dynamic feature clustering (CNNs&DFC). Finally, we validated the performance of these features on two publicly available datasets. Quantitative and qualitative experimental results on both intra-domain and inter-domain fashion clothing image retrieval showed that deep learning based feature representations far outweigh traditional hand-crafted feature representation. Additionally, among all deep learning based methods, CNNs with explicit feature clustering performs best, which shows feature clustering is essential for discriminative fashion feature representation.

Keywords: convolutional neural network, feature representation, image processing, machine modelling

Procedia PDF Downloads 109

Authors: Guezouli Larbi, Belkacem Soundes

Abstract:

The automation of the text detection process can help the human in his driving task. Its application can be very useful to help drivers to have more information about their environment by facilitating the reading of road signs such as directional signs, events, stores, etc. In this paper, a system consisting of two stages has been proposed. In the first one, we used pseudo-Zernike moments to pinpoint areas of the image that may contain text. The architecture of this part is based on three main steps, region of interest (ROI) detection, text localization, and non-text region filtering. Then, in the second step, we present a convolutional neural network architecture (On-Road Text Detection Network - ORTDN) which is considered a classification phase. The results show that the proposed framework achieved ≈ 35 fps and an mAP of ≈ 90%, thus a low computational time with competitive accuracy.

Keywords: text detection, CNN, PZM, deep learning

Procedia PDF Downloads 55

Authors: Fatma El-Zahraa El-Taher, Ayman Taha, Jane Courtney, Susan Mckeever

Abstract:

Understanding road networks plays an important role in navigation applications such as self-driving vehicles and route planning for individual journeys. Intersections of roads are essential components of road networks. Understanding the features of an intersection, from a simple T-junction to larger multi-road junctions, is critical to decisions such as crossing roads or selecting the safest routes. The identification and profiling of intersections from satellite images is a challenging task. While deep learning approaches offer the state-of-the-art in image classification and detection, the availability of training datasets is a bottleneck in this approach. In this paper, a labelled satellite image dataset for the intersection recognition problem is presented. It consists of 14,692 satellite images of Washington DC, USA. To support other users of the dataset, an automated download and labelling script is provided for dataset replication. The challenges of construction and fine-grained feature labelling of a satellite image dataset is examined, including the issue of how to address features that are spread across multiple images. Finally, the accuracy of the detection of intersections in satellite images is evaluated.

Keywords: satellite images, remote sensing images, data acquisition, autonomous vehicles

Procedia PDF Downloads 110

Authors: Emine G. Cansu Ergun

Abstract:

Conjugated system based sensors for selective detection of metal ions have been taking attention during last two decades. Fluorescent sensors are the promising candidates for ion detection due to their high selectivity towards metal ions, and rapid response times. Detection of mercury in an environmenet is important since mercury is a toxic element for human. Beyond the maximum allowable limit, mercury may cause serious problems in human health by spreading into the atmosphere, water and the food chain. In this study, a quinoxaline and 3,4-ethylenedioxy thiophene based donor-acceptor-donor type conjugated molecule used as a fluorescent sensor for detecting the mercury ion in aqueous medium. Among other various cations, existence of mercury resulted in a full quenching of the fluorescence signal. Then, a paper based sensor is constructed and used for mercury detection. As a result it is concluded that the offering sensor is a good candidate for selective mercury detection in aqueous media both in solution and paper based forms.

Keywords: Conjugated molecules , fluorescence quenching, metal ion detection , sensors

Procedia PDF Downloads 122

Authors: Pengchao Wei, Xiangzhong Fang

Abstract:

Noise is ubiquitous in this world. Denoising is an essential technology, especially in image semantic segmentation, where noises are generally categorized into two main types i.e. feature noise and label noise. The main focus of this paper is aiming at modeling label noise, investigating the behaviors of different types of label noise on image semantic segmentation tasks using K-Nearest-Neighbor and Convolutional Neural Network classifier. The performance without label noise and with is evaluated and illustrated in this paper. In addition to that, the influence of feature noise on the image semantic segmentation task is researched as well and a feature noise reduction method is applied to mitigate its influence in the learning procedure.

Keywords: convolutional neural network, denoising, feature noise, image semantic segmentation, k-nearest-neighbor, label noise

Procedia PDF Downloads 184

Authors: Eshta Ranyal, Kamal Jain, Vikrant Ranyal

Abstract:

Potholes are a severe threat to road safety and a major contributing factor towards road distress. In the Indian context, they are a major road hazard. Timely detection of potholes and subsequent repair can prevent the roads from deteriorating. To facilitate the roadway authorities in the timely detection and repair of potholes, we propose a pothole detection methodology using convolutional neural networks. The YOLOv3 model is used as it is fast and accurate in comparison to other state-of-the-art models. You only look once v3 (YOLOv3) is a state-of-the-art, real-time object detection system that features multi-scale detection. A mean average precision(mAP) of 73% was obtained on a training dataset of 200 images. The dataset was then increased to 500 images, resulting in an increase in mAP. We further calculated the depth of the potholes using stereoscopic vision by reconstruction of 3D potholes. This enables calculating pothole volume, its extent, which can then be used to evaluate the pothole severity as low, moderate, high.

Keywords: CNN, pothole detection, pothole severity, YOLO, stereovision

Procedia PDF Downloads 108