The section of preprocessing quite significant in terms of code because it facilitates the training of model of facial recognition and identify faces. First of all, the image is transformed into grayscale for reducing the size of data. In the mode of detection, the program is capable of detecting both eyes within the region of face. Once the eyes are detected, the distance present between the eyes is detected by the program and the whole face is scaled in such a way that distance remains the same. In addition to it, the eyes are adjusted in such a manner that they are at a specific height and are also horizontal. Lastly, an elliptical mask is placed on the detected face for cropping shadows or hair might cause issues in the identification of faces. When all of these processes are completed, the end result is a preprocessed face which has the dimensions of 70 by 70 pixels.

Initiation of Portable Low-Cost Facial Recognition

The user is capable of initiating training after faces have been collected. Moving on, the training set can be processed and initiated with the addition of new faces through the button of Add Person or it can be loaded through the use of Load Button. It is important for the training set to include different lighting conditions, angles, and expressions of faces for each and every person to be precise in recognition and identification. For instance, if faces are trained under the light which is strong on the face's right side while during the identification, the light is on the left side, it will have an adverse effect on the system and the result will be not satisfying. The objective is concerned with getting more variations between different faces in such a way that more information is saved and stored in the significant and principle components. Thus, the set of training must include different conditions for obtaining better outcomes.

Detection of Portable Low-Cost Facial Recognition

Through the use of eigenvalues and eigenvectors trained in the processes and model during the phase of training, the current preprocessed face is taken by the program and is projected into the subspace of PCA. Then, the projection is taken by it and the faced is reconstructed into an image. If the image of query is a part of the set which has been trained, the reconstruction would be easier. Moving on, the reconstructed image is compared by the program to the image which has been preprocessed through the use of an L2 relative calculation of error. For this error, a threshold is generally set and errors are calculated which are below this limit. These errors indicate the detected match.

Under predetermined distance and lighting conditions, the system of portable facial recognition is capable of achieving up to ninety percent validity and accuracy in the recognition of correct person. The lamp which was utilized had the intensity of 1800 lumen while 6500 Kelvin was the color temperature. From the webcam, 55 ± 5 cm distance was defined. In addition to it, the program is capable of accommodating approximately 6 individuals within the interface for recognition and training. When it comes to timing, two devices were utilized for the rate of recognition and detection. Raspberry Pi was the very first device in which the program was run. Besides this device, a benchmarking device was utilized and it was a Toshiba Computer. In comparison with each other, the latter was faster by the rate of 50ms. Meanwhile, the former was capable of going from capturing images to recognizing it within the 200ms average time. And anything less than one second was not perceivable to the user. Thus, Raspberry Pi was utilized as it was more suitable.

For the detection mode, there is no special action which is required. The program will be detecting the face automatically and a yellow rectangle will be put on the face while green circles will be put on the eyes. In this manner, the faces will be detected and identified.

For adding people, the button of Add Person had to be used and it served to add a person to the available training set. And each time a picture was taken, a white flash was exhibited for alerting the user. It is, however, important to understand that in the mode of Collect Face, there will be a red rectangle on the face of person who is training new faces. In the right edge of the interface, the most recent face will be displayed. If more faces have to be added to an individual who is already on the screen, the user will be required to click on that person’s most recent picture. This will commence the program of collecting faces for that individual.

Once the user finished the collection of faces, it was important for the user to click on a specific area in the interface which was not occupied by a button or face. Generally, this initiated the training of different preprocessed faces available in the model. And once the program was finished being trained, the program switched to the mode of recognition.

Overall, it can be said that the source code was written in the language of C++ as it was quite fast in comparison with MATLAB. Additionally, functionality was also added for saving and loading a number of faces which were previously trained. In the process of testing, the system was capable of achieving up to ninety percent precision in identifying the person who was trained under predetermined distance and lighting conditions. In addition to it, the system was capable of training the faces of six people. There was an issue in the process of training concerned with inefficient recognition. The process of recognition was sometimes not quick and it indicated the need of improving the algorithm. Therefore, this process can be further improved by adding a better system and enhancing the algorithm utilized in the system(Song, Kim, & Jeon, 2014).

Planning of Portable Low-Cost Facial Recognition

In order to achieve the success of this project, the planning was really fundamental. This chapter reviews complete planning along with the project management methods applied have

Project Management Techniques

Gantt Chart of Portable Low-Cost Facial Recognition

The workload of the project can be broken down into an accessible method by using a Gantt chart (Nurre & Weir, 2017). All of the tasks, as well as the sub-tasks, were added together as for the durations were. The tasks’ duration contained a startup date and also an expected date to complete the task. This results in the number of due days to every task, and the durations fixed on were those which appeared logically realizable.

Viewing the progress of the project in percentages can be done with Microsoft Project, whereas the percentages were highly valuable in observing the progression of the project itself. On the other hand, sometimes the project was not remaining on the track along with the Gantt chart

Version Control System or VCS of Portable Low-Cost Facial Recognition

A Version Control System or VCS – GIT (Blischak, Davenport, & Wilson, 2016)was used with a purpose to ensure that the project code will be safe and controlled. GIT has offered some valuable features to the code management, for example, such as restoring back the files to the earlier versions from the elder commits. The Incremental model will support this action to be highly useful since every single commit calculated as software build.

The ‘GitHub’ website was used with a purpose to keep the security of data and also make it possible to be accessed from any places. GitHub is stored inside the cloud by using a front-end website.

Time Spent on the Project of Portable Low-Cost Facial Recognition

The time spent for this project was mediocre of 16 hours a week. However, some weeks had slightly less time to be spent because of some deadlines for other parts of the coursework. In case if the development had begun former, then this might result in one or more features being installed into the end-product. 

Meetings with Supervisor of Portable Low-Cost Facial Recognition

A different perform that could be done for a part of this project was the meeting held with the supervisor. Over the timeframe of the project, only some minor helpful meetings held. However, in every single meeting, there are suggestions for the project along with the progress evaluation made by the supervisor, which were really helpful in completing this project. The main reason for these minor meetings with the supervisor was because of under pressure on fitting them with a personal schedule.

Planning and SLDC of Portable Low-Cost Facial Recognition

Even though it is not presented on the Gantt chart, there was a throughout the use of the Incremental software development type. Every task of development was its own software/build version. There were the re-visit of requirements, design, testing, and implementation for every Increment of software

Evaluation of Portable Low-Cost Facial Recognition

In this chapter, the Evaluation of this project will be discussed. This chapter will discuss some areas such as feedback along with the assumption of whether the aims/objectives/requirements have met.

There are three different areas were evaluated in this evaluation such as:

Evaluating if the objectives and aims of this project were achieved

Evaluating if the requirements have been accomplished

Evaluating and also observing the feedback from the users based on the questionnaire

Feedback of Portable Low-Cost Facial Recognition

The results from the questionnaire have provided insight considerations of the product:

All users thought that the idea was great and excellent

7/10 observed that the software is quite simple to navigate

8/10 stated that the performance of the camera was good

9/10 claimed that the camera produced excellent photos

In general, the users thought that the product had a unique design with a great face recognition feature

Requirements of Portable Low-Cost Facial Recognition

In order to evaluate whether the product’s requirements have been achieved or not, a table below has presented the evaluation of all the requirements in chapter 4.

Requirement	Type	Achieved/Evaluation
User Login	Functional	Yes – There was an implementation of the login
Image Capturing & Video Recording	Functional	Yes – the image capturing and video recording are completely implemented.
Data Saving	Functional	Yes – The implementation of the database is functioning with Python scripts.

Objectives Completed

The project has various objectives mentioned in chapter 1.2. The table below discusses whether all the objectives have completed or not.

Objective	Achieved?
Objective 1: Provide Security in Cheap Price	Yes – this portable face recognition system is low-cost
Making Portable	Yes – the design of this system is able to support the system to become portable
Recognition from different Angles	Yes – the UIs along with the software designs have made the product to be able in making recognition from different angles
Retrieve the basic information using API’s	Yes – all of the aspects such as requirements are discussed along with the usability and legal/ ethical issues that might be occurred

 Expectations of Portable Low-Cost Facial Recognition

The project has met the early expectations where all of the objectives were set out. Even though there were some hardware issues during completing this project, but the most essential and fundamental functions are integrated. Thus, there is a positive achievement since the requirements and objectives have been achieved gene