scholarly journals Real-time gaze estimation via pupil center tracking

2018 ◽  
Vol 9 (1) ◽  
pp. 6-18 ◽  
Author(s):  
Dario Cazzato ◽  
Fabio Dominio ◽  
Roberto Manduchi ◽  
Silvia M. Castro
Keyword(s):  
Real Time ◽  
Learning Algorithm ◽  
Natural Environments ◽  
Gaze Estimation ◽  
Head Pose ◽  
Data Set ◽  
Gaze Tracking ◽  
Illumination Changes ◽  
Wide Range ◽  
Estimation System

Abstract Automatic gaze estimation not based on commercial and expensive eye tracking hardware solutions can enable several applications in the fields of human computer interaction (HCI) and human behavior analysis. It is therefore not surprising that several related techniques and methods have been investigated in recent years. However, very few camera-based systems proposed in the literature are both real-time and robust. In this work, we propose a real-time user-calibration-free gaze estimation system that does not need person-dependent calibration, can deal with illumination changes and head pose variations, and can work with a wide range of distances from the camera. Our solution is based on a 3-D appearance-based method that processes the images from a built-in laptop camera. Real-time performance is obtained by combining head pose information with geometrical eye features to train a machine learning algorithm. Our method has been validated on a data set of images of users in natural environments, and shows promising results. The possibility of a real-time implementation, combined with the good quality of gaze tracking, make this system suitable for various HCI applications.

scholarly journals Alkalmazott mesterséges intelligencia felhasználási területei és biztonsági kérdései – Mesterséges intelligencia a gyakorlatban

2020 ◽  
Vol 1 (1) ◽  
pp. 35-42
Author(s):  
Péter Ekler ◽  
Dániel Pásztor
Keyword(s):  
Artificial Intelligence ◽  
Big Data ◽  
Anomaly Detection ◽  
Real Time ◽  
Learning Algorithm ◽  
Sensor Data ◽  
Test Field ◽  
Real Time System ◽  
Data Set ◽  
Wide Range

Összefoglalás. A mesterséges intelligencia az elmúlt években hatalmas fejlődésen ment keresztül, melynek köszönhetően ma már rengeteg különböző szakterületen megtalálható valamilyen formában, rengeteg kutatás szerves részévé vált. Ez leginkább az egyre inkább fejlődő tanulóalgoritmusoknak, illetve a Big Data környezetnek köszönhető, mely óriási mennyiségű tanítóadatot képes szolgáltatni. A cikk célja, hogy összefoglalja a technológia jelenlegi állapotát. Ismertetésre kerül a mesterséges intelligencia történelme, az alkalmazási területek egy nagyobb része, melyek központi eleme a mesterséges intelligencia. Ezek mellett rámutat a mesterséges intelligencia különböző biztonsági réseire, illetve a kiberbiztonság területén való felhasználhatóságra. A cikk a jelenlegi mesterséges intelligencia alkalmazások egy szeletét mutatja be, melyek jól illusztrálják a széles felhasználási területet. Summary. In the past years artificial intelligence has seen several improvements, which drove its usage to grow in various different areas and became the focus of many researches. This can be attributed to improvements made in the learning algorithms and Big Data techniques, which can provide tremendous amount of training. The goal of this paper is to summarize the current state of artificial intelligence. We present its history, introduce the terminology used, and show technological areas using artificial intelligence as a core part of their applications. The paper also introduces the security concerns related to artificial intelligence solutions but also highlights how the technology can be used to enhance security in different applications. Finally, we present future opportunities and possible improvements. The paper shows some general artificial intelligence applications that demonstrate the wide range usage of the technology. Many applications are built around artificial intelligence technologies and there are many services that a developer can use to achieve intelligent behavior. The foundation of different approaches is a well-designed learning algorithm, while the key to every learning algorithm is the quality of the data set that is used during the learning phase. There are applications that focus on image processing like face detection or other gesture detection to identify a person. Other solutions compare signatures while others are for object or plate number detection (for example the automatic parking system of an office building). Artificial intelligence and accurate data handling can be also used for anomaly detection in a real time system. For example, there are ongoing researches for anomaly detection at the ZalaZone autonomous car test field based on the collected sensor data. There are also more general applications like user profiling and automatic content recommendation by using behavior analysis techniques. However, the artificial intelligence technology also has security risks needed to be eliminated before applying an application publicly. One concern is the generation of fake contents. These must be detected with other algorithms that focus on small but noticeable differences. It is also essential to protect the data which is used by the learning algorithm and protect the logic flow of the solution. Network security can help to protect these applications. Artificial intelligence can also help strengthen the security of a solution as it is able to detect network anomalies and signs of a security issue. Therefore, the technology is widely used in IT security to prevent different type of attacks. As different BigData technologies, computational power, and storage capacity increase over time, there is space for improved artificial intelligence solution that can learn from large and real time data sets. The advancements in sensors can also help to give more precise data for different solutions. Finally, advanced natural language processing can help with communication between humans and computer based solutions.

scholarly journals Separating damage from environmental effects affecting civil structures for near real-time damage detection

2017 ◽  
Vol 17 (4) ◽  
pp. 850-868 ◽  
Author(s):  
William Soo Lon Wah ◽  
Yung-Tsang Chen ◽  
Gethin Wyn Roberts ◽  
Ahmed Elamin
Keyword(s):  
Damage Detection ◽  
Real Time ◽  
Environmental Effects ◽  
Environmental Conditions ◽  
Principal Component ◽  
Detection Methods ◽  
Real Time Monitoring ◽  
Civil Structures ◽  
Data Set ◽  
Wide Range

Analyzing changes in vibration properties (e.g. natural frequencies) of structures as a result of damage has been heavily used by researchers for damage detection of civil structures. These changes, however, are not only caused by damage of the structural components, but they are also affected by the varying environmental conditions the structures are faced with, such as the temperature change, which limits the use of most damage detection methods presented in the literature that did not account for these effects. In this article, a damage detection method capable of distinguishing between the effects of damage and of the changing environmental conditions affecting damage sensitivity features is proposed. This method eliminates the need to form the baseline of the undamaged structure using damage sensitivity features obtained from a wide range of environmental conditions, as conventionally has been done, and utilizes features from two extreme and opposite environmental conditions as baselines. To allow near real-time monitoring, subsequent measurements are added one at a time to the baseline to create new data sets. Principal component analysis is then introduced for processing each data set so that patterns can be extracted and damage can be distinguished from environmental effects. The proposed method is tested using a two-dimensional truss structure and validated using measurements from the Z24 Bridge which was monitored for nearly a year, with damage scenarios applied to it near the end of the monitoring period. The results demonstrate the robustness of the proposed method for damage detection under changing environmental conditions. The method also works despite the nonlinear effects produced by environmental conditions on damage sensitivity features. Moreover, since each measurement is allowed to be analyzed one at a time, near real-time monitoring is possible. Damage progression can also be given from the method which makes it advantageous for damage evolution monitoring.

Enhanced real-time head pose estimation system for mobile device

2014 ◽  
Vol 21 (3) ◽  
pp. 281-293 ◽  
Author(s):  
Euclides N. Arcoverde Neto ◽  
Rafael M. Duarte ◽  
Rafael M. Barreto ◽  
João Paulo Magalhães ◽  
Carlos C.M. Bastos ◽  
...  
Keyword(s):  
Real Time ◽  
Pose Estimation ◽  
Mobile Device ◽  
Head Pose Estimation ◽  
Head Pose ◽  
Estimation System

scholarly journals RGBD Based Gaze Estimation via Multi-Task CNN

2019 ◽  
Vol 33 ◽  
pp. 2488-2495 ◽  
Author(s):  
Dongze Lian ◽  
Ziheng Zhang ◽  
Weixin Luo ◽  
Lina Hu ◽  
Minye Wu ◽  
...  
Keyword(s):  
Large Scale ◽  
Image Synthesis ◽  
Position Estimation ◽  
Point Estimation ◽  
Head Pose Estimation ◽  
Depth Image ◽  
Eye Position ◽  
Gaze Estimation ◽  
Head Pose ◽  
Gaze Tracking

This paper tackles RGBD based gaze estimation with Convolutional Neural Networks (CNNs). Specifically, we propose to decompose gaze point estimation into eyeball pose, head pose, and 3D eye position estimation. Compared with RGB image-based gaze tracking, having depth modality helps to facilitate head pose estimation and 3D eye position estimation. The captured depth image, however, usually contains noise and black holes which noticeably hamper gaze tracking. Thus we propose a CNN-based multi-task learning framework to simultaneously refine depth images and predict gaze points. We utilize a generator network for depth image generation with a Generative Neural Network (GAN), where the generator network is partially shared by both the gaze tracking network and GAN-based depth synthesizing. By optimizing the whole network simultaneously, depth image synthesis improves gaze point estimation and vice versa. Since the only existing RGBD dataset (EYEDIAP) is too small, we build a large-scale RGBD gaze tracking dataset for performance evaluation. As far as we know, it is the largest RGBD gaze dataset in terms of the number of participants. Comprehensive experiments demonstrate that our method outperforms existing methods by a large margin on both our dataset and the EYEDIAP dataset.

Discriminative Robust Head-Pose and Gaze Estimation Using Kernel-DMCCA Features Fusion

2020 ◽  
Vol 14 (01) ◽  
pp. 107-135
Author(s):  
Salah Rabba ◽  
Matthew Kyan ◽  
Lei Gao ◽  
Azhar Quddus ◽  
Ali Shahidi Zandi ◽  
...  
Keyword(s):  
Dimensional Space ◽  
Accurate Estimation ◽  
Gaze Estimation ◽  
Head Pose ◽  
Facial Hair ◽  
Features Fusion ◽  
Higher Dimensional ◽  
Wide Range ◽  
Feature Information ◽  
Improving Accuracy

There remain outstanding challenges for improving accuracy of multi-feature information for head-pose and gaze estimation. The proposed framework employs discriminative analysis for head-pose and gaze estimation using kernel discriminative multiple canonical correlation analysis (K-DMCCA). The feature extraction component of the framework includes spatial indexing, statistical and geometrical elements. Head-pose and gaze estimation is constructed by feature aggregation and transforming features into a higher dimensional space using K-DMCCA for accurate estimation. The two main contributions are: Enhancing fusion performance through the use of kernel-based DMCCA, and by introducing an improved iris region descriptor based on quadtree. The overall approach is also inclusive of statistical and geometrical indexing that are calibration free (does not require any subsequent adjustment). We validate the robustness of the proposed framework across a wide variety of datasets, which consist of different modalities (RGB and Depth), constraints (wide range of head-poses, not only frontal), quality (accurately labelled for validation), occlusion (due to glasses, hair bang, facial hair) and illumination. Our method achieved an accurate head-pose and gaze estimation of 4.8∘ using Cave, 4.6∘ using MPII, 5.1∘ using ACS, 5.9∘ using EYEDIAP, 4.3∘ using OSLO and 4.6∘ using UULM datasets.

Description of a High-Profit Combination of Low-Cost Real-Time Survey Management Practices Used to Optimize Reservoir Landing in Unfamiliar Deep Offshore Geological Environment

2021 ◽  
Author(s):  
Fabien Momot ◽  
Marie-Jocelyn Comte ◽  
Chloé Lacaze ◽  
Anas Sikal ◽  
Efficience Balou ◽  
...  
Keyword(s):  
Real Time ◽  
Management Practices ◽  
Low Cost ◽  
Vertical Displacement ◽  
Uncertainty Reduction ◽  
Vertical Position ◽  
Reservoir Model ◽  
Data Set ◽  
Wide Range ◽  
Operational Time

Abstract After a first part of the drilling campaign, including about 10 wells and branches achieved within two years, the operator started questioning the geological reservoir model and reserves implications for the field Offshore Congo. Considering the potential economic impact of this development, the decision was made to reduce wellbore positioning uncertainty relying on optimization and survey QAQC processes that could be applied without adding cost of extra equipment, operational time or personnel. With more than 10 wells drilled using recent while drilling measurement and directional tools in the same environment, a wide range of wellbore positioning information was available for analysis, post-correction, and geological/reservoir model deeper understanding. Also, investigation was done to recover existing geomagnetic data acquired during the geophysical campaign. Thanks to this extensive data set, enhanced wellbores positioning was implemented using meticulous combination of processes. The "process" overall impact is often underestimated while most of the data is already available. For lateral positioning correction, it included the processing of geomagnetic IFR data over the Moho field associated to Multi Station Correction. For vertical repositioning, BHA sag correction was applied with scrutinous assessment of residual sag uncertainty and detailed analysis of continuous survey data. This robust, cost-effective, and valuable solution was chosen to be applied by the operator in the Moho field. The process was first applied post-drilling to evaluate the level of improvement that could be brought to another well also exposed to challenging trajectory context (ERD 2 with reduced target 25 × 50 m at almost 8000m MD/RT). It confirmed that the achievable uncertainty reduction would meet well objectives without adding any risk or operational time nor jeopardizing wellbore positioning and collision avoidance. Thus, it brought up to 50 to 60% of uncertainty reduction and about 30m lateral and 3m vertical displacement. The reduction of the uncertainty and trajectory adjustment allowed to enhance geologic context understanding. The vertical position of the well was offset following this revision. This had a 5% consequence in term of oil layer thickness for this well. Then, the team designed and rolled out to the operator and contractors an execution strategy and operational workflow including remote monitoring with near real-time survey QAQC that would ensure the best correction process customized for the specific drilling challenges. This monitoring enabled reducing the ellipsoid to ~20 by 50m radius at TD = 7618m. This allowed entering in the reservoir at the exact top of the structure, behind the fault that was the optimum in term of reserves and secured 90% of potential reserves of this well. The operator's choice of valuing the available information to enhance their asset is a very interesting way to optimize the past efforts put in wellbore positioning to face the current economically constrained environment.

scholarly journals Patch Based Multiple Instance Learning Algorithm for Object Tracking

2017 ◽  
Vol 2017 ◽  
pp. 1-7 ◽  
Author(s):  
Zhenjie Wang ◽  
Lijia Wang ◽  
Hua Zhang
Keyword(s):  
Object Tracking ◽  
Real Time ◽  
Learning Algorithm ◽  
State Of The Art ◽  
Multiple Instance Learning ◽  
Partial Occlusion ◽  
Learning Rates ◽  
Classification Score ◽  
Object Based ◽  
Illumination Changes

To deal with the problems of illumination changes or pose variations and serious partial occlusion, patch based multiple instance learning (P-MIL) algorithm is proposed. The algorithm divides an object into many blocks. Then, the online MIL algorithm is applied on each block for obtaining strong classifier. The algorithm takes account of both the average classification score and classification scores of all the blocks for detecting the object. In particular, compared with the whole object based MIL algorithm, the P-MIL algorithm detects the object according to the unoccluded patches when partial occlusion occurs. After detecting the object, the learning rates for updating weak classifiers’ parameters are adaptively tuned. The classifier updating strategy avoids overupdating and underupdating the parameters. Finally, the proposed method is compared with other state-of-the-art algorithms on several classical videos. The experiment results illustrate that the proposed method performs well especially in case of illumination changes or pose variations and partial occlusion. Moreover, the algorithm realizes real-time object tracking.

scholarly journals Scalable and Reliable Deep Learning Model to handle real-time Streaming Data

2020 ◽  
Vol 9 (3) ◽  
pp. 3840-3844 ◽  
Keyword(s):  
Deep Learning ◽  
Real Time ◽  
Data Streams ◽  
Learning Algorithm ◽  
Streaming Data ◽  
Experimental Result ◽  
Time Data ◽  
Data Set ◽  
Public Data ◽  
Real Time Data

We have real-time data everywhere and every day. Most of the data comes from IoT sensors, data from GPS positions, web transactions and social media updates. Real time data is typically generated in a continuous fashion. Such real-time data are called Data streams. Data streams are transient and there is very little time to process each item in the stream. It is a great challenge to do analytics on rapidly flowing high velocity data. Another issue is the percentage of incoming data that is considered for analytics. Higher the percentage greater would be the accuracy. Considering these two issues, the proposed work is intended to find a better solution by gaining insight on real-time streaming data with minimum response time and greater accuracy. This paper combines the two technology giants TensorFlow and Apache Kafka. is used to handle the real-time streaming data since TensorFlow supports analytics support with deep learning algorithms. The Training and Testing is done on Uber connected vehicle public data set RideAustin. The experimental result of RideAustin shows the predicted failure under each type of vehicle parameter. The comparative analysis showed 16% improvement over the traditional Machine Learning algorithm.

scholarly journals Low Cost Eye Tracking: The Current Panorama

2016 ◽  
Vol 2016 ◽  
pp. 1-14 ◽  
Author(s):  
Onur Ferhat ◽  
Fernando Vilariño
Keyword(s):  
Eye Tracking ◽  
Visible Light ◽  
Open Source ◽  
Low Cost ◽  
Gaze Estimation ◽  
Head Pose ◽  
Gaze Tracking ◽  
Estimation Techniques ◽  
Public Datasets ◽  
Interesting Alternative

Despite the availability of accurate, commercial gaze tracker devices working with infrared (IR) technology, visible light gaze tracking constitutes an interesting alternative by allowing scalability and removing hardware requirements. Over the last years, this field has seen examples of research showing performance comparable to the IR alternatives. In this work, we survey the previous work on remote, visible light gaze trackers and analyze the explored techniques from various perspectives such as calibration strategies, head pose invariance, and gaze estimation techniques. We also provide information on related aspects of research such as public datasets to test against, open source projects to build upon, and gaze tracking services to directly use in applications. With all this information, we aim to provide the contemporary and future researchers with a map detailing previously explored ideas and the required tools.

Sign in / Sign up

Export Citation Format

Share Document