FACE DETECTING SYSTEM FOR USING IN SAFETY CIRCUMSTANCES

Student Name

Institute Name

Abstract

The report will discuss in detail about the capstone project that proposes a structure that will be able to solve certain gap in security in constructing by implementing facial recognition as well as IoT technology. The issue here is that it deals with constant crowd focused management, emergency event track in a big venue as well as dealing with number of inconsistencies along with human mistakes that are caused purely based on current allocation of wardens. This shows a weakness in timing, interaction and further restricts sources of security to particularly those are part of assignment. This report discusses in detail the overall scope of the project followed by the methods as well as designing inclusive research procedure which are utilised along with breaking down of stages.

Contents

Abstract 2

Introduction 4

Explorative research 6

Design steps 6

Confirmative research 7

Research result 7

Design focused generalization 7

Design Novelty 7

Prototype 8

Application 8

Project restrictions 8

Ethical as well as legal considerations 8

Literature review 9

Contextual usage 9

Review of answers 10

Android attendance structure 10

Face API door-lock 11

Open CV automated and lock 11

Smart Campus Video Footprint 12

Attendance of classroom 14

Convolutional Neural Network also called as CNN 15

Dataset for training 15

Facial recognition modes 15

Hardware Needs 16

Camera 16

Computation power 16

Conclusion 16

Reference 18

Introduction

In an event of emergency, an important concern is based on the execution of security as well as emergency steps is human mistake, since the circumstances are normally high-press as well as may leave some space for human mistake. Developing security structure with this in mind, along with some clarity over the inept or wrong psychology permits for some innovation that rightly able to fill the gaps causing and inherent finishing shareholders requirements. An instance of this case is at the time of fire emergency cases in a big building, where fire executives can be part of allocation to answer suddenly and make sure that every patron have left the developed industry, pose a rising challenge on themselves.

The research assignment as well as POC also called as proof of concept focus to further improve the security of constructing and different occupants, to be utilised in an education-based facility. This, in the end, meets the gap by supervising and testing an industry a department where people can enter or exit the place.

Current security structure is normally in two parts, conventional surveillance cameras as well as smoke detection fire focused alarms, such technology need a lot of communication to drive any correct kind of data or information. The primary intention is specifically for local alarms along with manual history focused footage reviews, this cannot be high advantageous ensuring any emergency for number of wardens as well as other security managers. In order to meet this challenge, the designed structure permits for constructing a network related with camera as well as sensors which can be automatic test crow’s movement along with environment information, in turn offer valued data (Cuimei et al., 2017). The overall analyses of such data permit for security managers in order to track the overall origin as well as present location in case of an emergency, finding some patrons which can be stuck as well as deliver automatic alerts as well as reports. The sudden gain of this can be relieving personnel and constructing workforce off major pressure and options to make an error and behave as a physical form of alarm, with visual as well as audio options. In addition, gains consist of a live interface for understanding the incoming information, history focused databases, communication system location maps as well as prospective for detailed data focused manipulation for offering added data external of emergency cases.

An instance of this operational is at time of fire that break down in an industry of a construction. Sensors can diagnose this as well as may trigger a physical alarm, as well as simultaneous alerting staff with valued data like location, count of crowd and time. Facial identification or recognition can also count the people and their number which have entered the construction, permitting security personal to constantly identify life-threating issues (He et al., 2017). In addition, it tests important literature, grows artefacts and manufactures architecture consisted for the gain of developing information and knowledge. The overall architectures which are consisted cover definition, system, protypes as well as other ways in other to offer additional overviews of diverse features and showing some kind of evolution of the different designs. There is an attention on diverse patterns along and utilises situation to focus on many gains as well as artefacts application. The whole system will consist a network focused on cameras, microprocessors, sensors along with deep understanding to develop a correct kind and functional answers. Part of such design will further be prototyped in a small answer, that will function in the form of proof of definition and accomplish some of the architectural designing that will be discussed or elaborated. The software can be generic and designed to add some utility of the hardware execution. The developed information can further be executed in number of applications, since the methods keep generalized along with novelty in mind at the time all important kind of decisions.

At the same time, safety and wardens manage to access the dashboard for the data enabled them to further react as required. There are number of considerations that require to be taken into focus for the execution of such structure, these consist financial limitations, legal issues, ethical issues as well as the motive of the whole research.

Research methodology: DIR also called as Design Inclusive Research that come from an epistemological angle, is a methodology that focuses on design a mid-level research that need growing strong theory, it offers data on high level, based on context as well as integration. In addition, with the help of procedure of design is developed that merge both design as well as research. The gain of this principle is the capacity to construct new information dependent on context utilising procedure and artefacts. Artefacts here is based on creations developed at the time such projects, consisting software and hardware (Jain and Patel, 2016).

Figure: Phases of DIR

The design methodology can further be break down into 3 kinds of methodology stages:

Explorative research, design-based steps and confirmative focused research. These can be additionally understood in the current graph.

This way, research is further forced to develop many kinds of models, offer help for the concepts via validation methodology as well as to construct an exceptional clarity of the current study. The benefits as a whole are that it constructs information and realize many kinds of disciplines. The most significant result of the DIR principle is knowledge development, this is about diverse features which may be in the form of architecture, descriptions as well as varied patterns.

Explorative research

This stage consists of researching present information and consisting the found data to analyse the issue. This way, a clarity and criteria can be eliminated to further breakdown present methodology for permitting the growth of the theory and hypothesis along with the whole design. This was done in detail at the time of the first element of this assignment, the research preparation where the present data was simple as well as compared, along with its significant along with intended answers.

Design steps

The next stage in DIR is the stage of design steps or actions, which further divided into three parts like:

Conception, prototypes and designs. These elements consist of developing models and methodology and critiquing them in order to lead a prototype kind of decisions. This procedure will permit for methodical evaluating concepts as well as models to divide between correct and viable answers. This is the angle of this assignment present in architectural and design methods considerations as well as alternate options along with many breakdowns (Kazanskiy et al., 2017). It is further followed by testing with certain prototypes following elements of selected designs.

Confirmative research

This end phase of the DIR procedure is the process of testing decisions and results, validating methodology and principle while testing the outcomes of growth. This way, it is also possible to decide the generalization and novelty of the artefacts designed.

This stage is crucial in clearing the whole project end results and gaining clarity over what was actually possible related with project. This way of the project can be altered and diversion that come from starting hypothesis may have happened. This can be a side-effect of design principles, where certain deviation as well as evolution of studies as well (Kim and Wei, 2017).

Research result

Design focused generalization

The process of generalization is developing research is for a system which is not just implemented to the particular case it is being tested on, however rather can be correct in diverse cases as well as contexts. To gain success in growing a well-generalised development, patterns should be implemented to many cases. The improved information must, in a way, be followed this concept and thus impacting designs. This is under a category utilising the DIR methods, since the principle of making sure the success of the design as exhibited as part of consolidation, in the process of generalization (Kaziakhmedov et al., 2019). It is also correct to gain clarity over the procedure of research to develop new kind of data or knowledge. The design is about the process that allow the methodically testing of ideas and then handling the viable and appropriate solutions. It comes with the confirmative research with the DIR process that includes the procedure which is based on verifying on decisions and outcomes. It includes the validation of methods and the approaches through analysing about the results of development. It is possible to determine about the novelty and the generalization with the artefacts. This stage is important with clarifying about the project on final outcome and outstanding context of project. The direction of project might be altered with diversion from the initial hypothesis.

Design Novelty

The concept of design novelty is based on the newness of data developed, where there are number of improvements of present research. The concept of Novelty can be associated to the invented discovered information, based on the context of its growth. Research in present facial recognition structures as well as fire alarms presently grows isolated information as well as artefacts, thus the present research offered by novelty because of the improvement in the present knowledge and the development of data as well as artefacts which does not exist here (Kumar et al., 2019). The novelty design is based on the research for the current facial recognition system that includes the current development of the isolated growth and the artefacts. It is about research providing due to the improvement of knowledge and creation of information and artefacts that are not present.

Prototype

Prototype is considered as proof of definition of different design as well as research is an important result to offer malleable proof of the viable execution, along with the gain of permitting for theory-based designs to show realistic answers. The prototype consists the important system functions and needs, also offering a platform for additional research by number of other parties. This step is manufactured at the time of second stage of the DIR methods, since it will help confirmative research as well as back the project motive.

Application

Such kind of application this developed can be majorly broader, at the same time the study is references a particular context. As needed in design focused generalization, the design is able of being adapted for many contexts and motives. Focusing on the system itself executes elements and abilities which are not context particular nor it is sourced from bespoke kind of hardware, it permits for other to execute the same and comfortably implement the research.

Project restrictions

The projects were under the restriction of personal finance limitation and pre-decided deadlines. These elements also play a part in restricting the project scope and however was ample for growth of a proof of research.

Ethical as well as legal considerations

A structure which is capable of identifying faces brings forward many things which must be focused as per the ethical and legal issues. Ethical challenges, evaluating systems in the public culture in reduced contextual since it normally an international issue, something firms like WHO also called as World Health Organization have developed directions. This can need signage, a discussion and paperwork on the kind of data is being saved and utilised. this way, the legal needs of such technology should be researched for additional development here (Lu et al., 2017). In Australia, the Australian Privacy Act in the year 1988, along with different government policies, setup laws which cover a lot of technology that can impact people’s information as well as privacy. Such act particularly discusses the collection, use, integrity and correct kind of personal data, along with responsibility of data and accessing to an individual’s records. This is crucial to gain clarity over the contextual research, since it can possibly invoke some punishment from government firms in case complaints are registered.

The main concerns which are considered are the challenges of right kind and depending on an automatic detection structure, use of data externally of intended motive and incorrect aggregation of data leading to concepts or assumptions (Kurnianggoro and Jo, 2019). The present mitigated strategy comes with lack of identification in video supervision, and depending just on detecting faces, no storage of images that are captured at the same time such operational is offered for demo motive only and limited access towards live footage.

At the same time, such strategies are not ideal, it is identified which is ample for the evidence of concept, as well as execution of the end product will require additional research in ethical as well as legal direction, similar to warning signatures and publicise the usage of data.

The primary factors are related to the consideration about the issues related to the correctness which comes on the automation about the detection system. It is about the use of information with inappropriate aggregation for the data that can lead to certain assumptions. It comes with no storage of data. Hence, the government guidelines includes the laws with covering the technologies with effect on the people privacy and information. It comes with accountability of information with access to the records of person. The strategies are not perfect with the proof of concept with the final product and then research is about the legal and the ethical guidelines like the warning signs.

Literature review

Contextual usage

Research on the significance of the gains of this technology, normally, offered a clarity that there is a clear requirement and gain to its growth. It works in a non-intrusive for with security technology, that offers techniques for improve the effectiveness, identifying kinds of personnel or occupants in a constructure and supervising human presence. In the research of one thousand individual, it was exhibited that sixty four percent of them believed which facial identification structures must be implanted for the security gains. Such facts are helped by the clarity of human mistake in situations that can be life threatening or can create some stress (Kushsairy et al., 2016).

Review of answers

System of same utilisation have been executed in the past as well as during the current time frame. Each methodologies principle is the answer in significance to the context, also, there is data to select each for this utilised case.

Android attendance structure

An android solution is considered as one that executes the present software in an Android phone, permitting students to further open a certain application that will utilise the front facing camera of the phone as well as scanning of face. This permits students to further confirm the attendance at every class has a developed QR code, as well as the phone of lecturer can get an updated list of students. In the end, a Raspberry Pi 3 model B+ is utilised in the form of server to process data. This case is used to gain clarity over the versatile of potential structure. The server utilises a Viola-Jones algorithm in order to crop certain faces from certain images, then it can be resized as well as convert to grayscale, where each pixel is utilised in the form of vector. 3 classifiers were part of comparison and the elements are LR also called as logistic regression, K-near Neighbour, Linear Discriminant analysis also called as LDA for the low-level resources as well as computational needs (Mady and Hilles, 2017). The outcomes of such study compared the gains of training every kind of classifiers, the processing strength or power, as well as the correctness. It exhibited that K-NN gains most from learning based datasets where 2.31 percent rise in accuracy utilising a thirty percent rise in training data at the same time also have the lowest level of training time (60.96s present at fifty percent data as well as 52.99s present at eighty percent information). At the same time, LR has the biggest starting correctness with 93 percent, it is very heavy and slow as well as well as never gain as much any kind of training. The understanding is based on the server making use of the Viola Jones algorithm which is for handling the crop of faces from the image It is then resized and then converted to the grayscale. It is for the results of the study which comes with the training as well.

Face API door-lock

Utilising the FACE API, a smarter door offering biometric important security was executed. This structure further controls a solenoid for the door locked methods, executed a PI3-B+ for the motive of processing as well as a Microsoft HD live Webcam for the facial identification. The FACE API operates on the Windows 10 IOT operational structure as well as an Azure level of subscription. This kind of establishment is also capable to offer audio feedback on the outcome of the scan for successful or failing scans. The most important gain of this procedure is utilising of Azure deployed platform for GPU help and Neural network execution, were similar to UTS usage case study a termed set of pictures (Niu and Chen, 2018). Such study is utilised for the understanding how such structure may further be utilised to supervise doors and execute FACE (Rajesh and Naveenkumar, 2016).

Open CV automated and lock

In other studies, facial identification is utilised to automatic and supervise a home for the elder people. The attention was to have a comfortable to utilise some interface, utilising an Android application to utilise the automated characteristics and accessing the door cameras. The server is executed utilising a PI3-B+ for processing all the data as well as operating automated function. The PI3 in itself is programmed utilising Python as well as the OpenCV API for many facial identifications as well as the API can be fed utilising images of called people for number of credentials. One can activate in case when some individual press the doorbell or when operations can be detected utilising a background subtracted algorithm. After grabbing such image, the face detection as well as segmented from the picture utilising Haar characteristics, this is then part of comparison of the database. PCA also called as Principle Component analysis algorithm as well as SVM also called as Support Vector machineries is utilised for some recognition. Such recognition is part of benchmark as well as trained utilised the Yale database, utilising PCA to get characteristics as well as SVM classifier is part of training (Mady and Hilles, 2017).

The test of benchmarks as part of FAR also called as False Acceptance rate as well as False rejection ratio, where the less value, the better. This structure exhibited that at a sixty percent threshold as well as thirty characteristics FAR and FRR were fifty three percent and 1.4%.

Smart Campus Video Footprint

A foundation called as Deep Guide executes to develop a video trajectory of institute with a footage archived using present camera structure and decreasing time to further find a particular target in the camps alleviate review time with an important footage. The experiments here prove that it can accomplish high level of correctness for number of students not close enough for the surveillance kind of cameras.

Camera can detect operation as well as record certain footage, then faces are part of detection and identified utilising Fisherface or CNN face identification. The structure than easily maps the centre pixel of mad maps as well as face, it is for the right location of the surveillance phase, the student ID as well as location is also recorded. The correctness of face recognition can be improved utilising CNN to execute Deep level learning.

The CNN architect utilised 3 convolutional form of layers extract fine grained characteristics, 1 flat layer that convert the two-dimension matrix to one-dimension results, three complete connected layers measuring the possibility to match face recognizers as well as one normalized layer measuring the overall probability to match facial identifiers as well as one normalised layer utilising the softmax operations. To build the database faces, every student utilised the Deep Guide applications in order to take some fronts as well as side opinion photos of people (Savita et al., 2018).

The face processing can be done via laptop operating Window ten, communication with current server utilising application of IEEE 802.11 access elements. It is operating an Open CV and can google Tensor flow libraries utilising a JAVA GUI. This can be utilised to scrub some footage in the situation of emergency or case to highlight people as well as locations.

To further benchmark or supervise the structure, the CNN as well as Fisherface depending recognition were evaluated for detected correctness, FPR or False Positive rate and implementation time in brighter, typica and dark light as well which can be repeated 10 times as well as average as part of distances from zero meters to five meters in 0.5 m case intervals.

These were part of training utilising the open face form of database PICS. It is part of realization where the correct and implemented time reduces majorly after three meters as well as many consecutive foundations that can cause high level of FPR. It is also part of conclusion after 3 meters and many connecting consecutive frames which can cause high level of FPR. One can also conclude where the CNN architecture was a lot more correct than Fisher dependent on all cultures, this is primarily correct in cases where a person is far from the camera also ninety percent accuracy when compared to seventy percent. CNN is also advantageous in that the implementation time never raise while improving the facial recognition correctness (Savita et al., 2018). One instance is utilised to assist a student to find a lost kind of wallet, where overall average time spent for three consumers to scrub the footage as well as find it as part of 8.32 minutes utilising conventional footage and 2.29 minutes utilising Deep Guide, since just footage with the importance face detected is utilised.

Attendance of classroom

This principle execute R-CNN also called as Region-CNN for targeted detection as well as SeetaFace API for facial recognition algorithm utilising the Wider Face data set. This structure implements a two-phase network, one for such network as well as other object detecting technology and the other especially for object detection, share a volume base layers for computer and ideal gains. The concept of SeetaFace includes of 3 models:

Face detection, face characteristics element placing and facial recognition. The first model executes a shallow kind of network funnel cascade architecture which is complemented by characteristics pointing and can solve non-linear map challenges, the correctness of every of this then is improved additionally utilising an ANN for characteristics extracting as well as relative features (Sharma et al., 2019).

Such study tests the effectiveness of diver video qualities at up to ten meters. The seven hundred and twenty p camera characteristics element which is at <4 meters, 1080p at <6 m and 4k at 9m. As part of 4k footage, the small individual face is 61*61 px at the same time the biggest is considered as 153*153 px. The structure in the end can be broken down into 5 parts:

HD video camera, distinct frame picture model, rapid R-CNN to diagnose faces and make some coordinated for people, SeetaFace for identification as well as automated attendance model. These will operate at the time of marketing indictors that is absent, late arrival, early time departure, carelessness and random level access.

In the end, it was determined that R-CNN with Seetaface mixed with a 4k Camera provide a satisfactory and correct outcome.

Convolutional Neural Network also called as CNN

Normally, executing recognition focused algorithm has two sections: representation dependent methodology to convert two-dimension pictures to another level of space to permit statistics processes to analyse certain kind of patterns e.g. SVM and FisherFace along with feature based processes to get characteristics then utilise a classifier for utilising these also called as CNN literature reviews have all showcased the significance of executing CNN to increase the correctness of recognition as well as detecting features. This is motivated with the help of biology as well as in the initial few kind of layers consist a convolutional level layer with pooled cascade to further simulate cascade of cells for characteristic extraction. Every convolutional neuron gets as well as also reacts to input from the past layer (Surinta and Khruahong, 2019). The latter layers are normally the external layers.

Dataset for training

There are number of database present for training in CNN to raise correctness and effectiveness. These variety present in complication and sizes. This transaction is normally resourced in heavy manner in one instance pushes GPU use for zero percent to ninety nine percent. There are number of these are present: ORL (four pictures of forty people), wide face dataset, database from Yale, Feret Database as well as self-developed dataset from consumers as well as students.

Facial recognition modes

It is important to utilise the most crucial model for the structure for the neural form of network to execute. A relative form of these is present in Face Recognition system in figure eleven. These API’s offer diverse gains, however other like SeetaFace as well as OpenCV must also be part of comparison.

Hardware Needs

Camera

The high number of live pixels, the greater detailed of detection in one instance, a four-k camera made an important different when compared to 1080p or less. At the same time, this will be based on the models as well as architecture utilised, since there are instances executes small sensor that 1080p or even less which covers Microsoft HD live, Android Front facing camera as well as Pi Camera (Savita et al., 2018). The fundamental challenge in this kind of considerations is the expense of hardware since 4k cameras are normally the costliest and less available.

Computation power

There are number of literature reviews use a Raspberry Pi 3 Model B+ in case of processing core ARM cortex A53 while many other utilise standard CPUs like Intel 17 CPU windows laptop. Another experimentation analysis the gain of diverse graphics cards for processing this data.

The gains of utilising a GPU despite is are when it offers multiple core parallel computing as well as has a massive number of cores, fast memory access as well as high level of floating-point computed power. Another gain is utilising of Nvidia’s CUDA as well as CUDA Deep Neural Network. Power from GPU varies majorly based on the use scenario, in one evaluation can be seen where high-end Desktop GPU has a major gain over some kind of Machine Learning GPU’s (Wang et al., 2017).

Hardware Setup

The Raspberry PI hardware setup is for the prototyping which is simple. Hence, the PI is connected to the micro USB power source with the camera ribbon which is for the camera ribbon to put in.

Software Setup

This is for the setup and the configuration for the tedious process. It is for set up to the proper environment with handling the research of the previous experiences. It includes:

1. The installing of the fresh full Raspbian
2. Installing the OpenCV and dependencies.
3. Retrieving the OpenCV library and modules
4. Handling the compiling of the library with the installation of camera modules and then configuring the python scripts for the capture of images.
5. The configuring about the Node Reds Watson IoT connection and then handling setup of the appropriate nodes with local and cloud.

The Open CV is open source which is for the image recognition library with the BSD licensing that is used for the contexts which is for providing the trained models. Hence, there are different options for the methods of detection and then they are also over 2500 algorithms. The library also has the platform which is extensive with the interface support like Python and Java, Windows and Linux. The benefit of the ease of distribution with the international collaboration with the extensive documentation with ease of adaptation with the development on future improvement. The API is for the concept which is used for the multiple API which is integrated for the completion of solutions. The API is for the implementation through the Node Red modules and the python. Hence, the functionality is about the performing the different ideal functions that are for the viable designs with handling the database storage and logging. It is about the dashboard and the view of the live charts and statistics with the view of the database of sensors.

Conclusion

The fire alarm system of face detection seems to show a very operational merge between two kind of technologies into modern kind of IoT answer. It is highlighted to be advantageous to have consistent access to data which is being developed by the system as part of demonstrations. The developed level of data deal with an architecture issues, where many patterns should cover to formulate the completely functional end product. Here, a description of every identified pattern along of the whole system architecture is provided to manufacture the functional evidence of concept. Such designs have the intention to help future growth by permitting for work on personal elements and a clarity of the communications. The design of this study or system is part of intention to permit for scalable and general use of the goods, and an adapted and viable answer which can be comfortably replicated as well as can be improved (Wati and Abadianto, 2017). The significant part, growth of this good is accessed because of the execution of open source libraries as well as comfortable to utilise API. The whole project focused to grow a structure that is able of identifying human facial features to further track movement of crowd as part of venue. this will offer important data to security personnel and workforce for gaining some clarity of the locations of patrons at the time of any emergency event an interpolation of data focused analysis of the past events. The operational feature of this in addition further improve by the added sensor for supervising the culture simultaneous manner, offering a method to trigger alarms in situation of a fire and track dramatic level of transformations in such variables (Yang and Jiachun, 2018). The alarms further cover a broader variety of mid-level consisting text messages, dashboard and emails and many more. The gain of many variance of many notifications is based on permitting for many people as notification, identifying that in some situations a text message may never be correct or viable for workforce and can be similar as emails.

Reference

Cuimei, L., Zhiliang, Q., Nan, J. and Jianhua, W., 2017, October. Human face detection algorithm via Haar cascade classifier combined with three additional classifiers. In 2017 13th IEEE International Conference on Electronic Measurement & Instruments (ICEMI) (pp. 483-487). IEEE.

He, Z., Zhang, J., Kan, M., Shan, S. and Chen, X., 2017. Robust fec-cnn: A high accuracy facial landmark detection system. In Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition Workshops (pp. 98-104).

Jain, V. and Patel, D., 2016. A GPU based implementation of robust face detection system. Procedia Computer Science, 87, pp.156-163.

Kazanskiy, N., Protsenko, V. and Serafimovich, P., 2017. Performance analysis of real-time face detection system based on stream data mining frameworks. Procedia Engineering, 201, pp.806-816.

Kaziakhmedov, E., Kireev, K., Melnikov, G., Pautov, M. and Petiushko, A., 2019, October. Real-world attack on MTCNN face detection system. In 2019 International Multi-Conference on Engineering, Computer and Information Sciences (SIBIRCON) (pp. 0422-0427). IEEE.

Kim, K.C. and Wei, H.T., 2017. Development of a Face Detection and Recognition System Using a RaspberryPi. The Journal of the Korea institute of electronic communication sciences, 12(5), pp.859-864.

Kumar, A., Kaur, A. and Kumar, M., 2019. Face detection techniques: a review. Artificial Intelligence Review, 52(2), pp.927-948.

Kurnianggoro, L. and Jo, K.H., 2019, November. Attention-Guided Model for Robust Face Detection System. In Pacific-Rim Symposium on Image and Video Technology (pp. 41-51). Springer, Cham.

Kushsairy, A.K., Kamaruddin, M.K., Nasir, H., Safie, S.I., Bakti, Z.A.K., Isa, M.R. and Khan, S., 2016, May. Embedded vision: Enhancing embedded platform for face detection system. In 2016 IEEE International Instrumentation and Measurement Technology Conference Proceedings (pp. 1-5). IEEE.

Lu, J., Fu, X. and Zhang, T., 2017, August. A smart system for face detection with spatial correlation improvement in iot environment. In 2017 IEEE SmartWorld, Ubiquitous Intelligence & Computing, Advanced & Trusted Computed, Scalable Computing & Communications, Cloud & Big Data Computing, Internet of People and Smart City Innovation (SmartWorld/SCALCOM/UIC/ATC/CBDCom/IOP/SCI) (pp. 1-4). IEEE.

Mady, H.H. and Hilles, S.M., 2017. Efficient Real Time Attendance System Based on Face Detection Case Study “MEDIU Staff”. International Journal on Contemporary Computer Research (IJCCR), 1(2), pp.21-25.

Niu, G. and Chen, Q., 2018. Learning an video frame-based face detection system for security fields. Journal of Visual Communication and Image Representation, 55, pp.457-463.

Rajesh, K.M. and Naveenkumar, M., 2016, December. A robust method for face recognition and face emotion detection system using support vector machines. In 2016 International Conference on Electrical, Electronics, Communication, Computer and Optimization Techniques (ICEECCOT) (pp. 1-5). IEEE.

Savita, K.S., Hasbullah, N.A., Taib, S.M., Abidin, A.I.Z. and Muniandy, M., 2018, November. How’s the Turnout to the Class? A Face Detection System for Universities. In 2018 IEEE Conference on e-Learning, e-Management and e-Services (IC3e) (pp. 179-184). IEEE.

Sharma, R., Ashwin, T.S. and Guddeti, R.M.R., 2019. A novel real-time face detection system using modified affine transformation and Haar cascades. In Recent Findings in Intelligent Computing Techniques (pp. 193-204). Springer, Singapore.

Surinta, O. and Khruahong, S., 2019. Tracking people and objects with an autonomous unmanned aerial vehicle using face and color detection. In 2019 Joint International Conference on Digital Arts, Media and Technology with ECTI Northern Section Conference on Electrical, Electronics, Computer and Telecommunications Engineering (ECTI DAMT-NCON) (pp. 206-210). IEEE.

Wang, F., Chen, X., Wang, D. and Yang, B., 2017, July. An improved image-based iris-tracking for driver fatigue detection system. In 2017 36th Chinese Control Conference (CCC) (pp. 11521-11526). IEEE.

Wang, Z. and Gao, F., 2017, March. An embedded parallel face detection system based on multicore processor. In 2017 IEEE 2nd Advanced Information Technology, Electronic and Automation Control Conference (IAEAC) (pp. 2684-2688). IEEE.

Wati, D.A.R. and Abadianto, D., 2017, November. Design of face detection and recognition system for smart home security application. In 2017 2nd International conferences on Information Technology, Information Systems and Electrical Engineering (ICITISEE) (pp. 342-347). IEEE.

Yang, W. and Jiachun, Z., 2018, July. Real-time face detection based on YOLO. In 2018 1st IEEE international conference on knowledge innovation and invention (ICKII) (pp. 221-224). IEEE.

The post FACE DETECTING SYSTEM FOR USING IN SAFETY CIRCUMSTANCES appeared first on My Assignment Online.