1. Trang chủ
  2. » Luận Văn - Báo Cáo

Khóa luận tốt nghiệp: Facial recognition technology in gym management.

54 0 0
Tài liệu đã được kiểm tra trùng lặp

Đang tải... (xem toàn văn)

Tài liệu hạn chế xem trước, để xem đầy đủ mời bạn chọn Tải xuống

THÔNG TIN TÀI LIỆU

Thông tin cơ bản

Tiêu đề Face Recognition for Gym Management System
Tác giả Nguyen Van Thanh Duc
Người hướng dẫn Do Trong Hop, Dr.
Trường học University of Information Technology
Chuyên ngành Information Systems Engineering
Thể loại Graduation Thesis
Năm xuất bản 2023
Thành phố Ho Chi Minh City
Định dạng
Số trang 54
Dung lượng 29,99 MB

Cấu trúc

  • CHAPTER 1. INTRODUCTION 0-5 (8)
    • 1.1. Problem Statement 00... ccc (8)
    • 1.2. Problem Solution 8n (8)
    • 1.3. Challenges (9)
    • 1.4. Goal and Study 2v h3 (0)
    • 1.5. Thesis SÍTUCfUTG................... ..- 2. 1S nh Tho Thọ TH Tu TH HH HH HH nh TT 8 (0)
  • CHAPTER 2: LITERATURE REVIEW..................... LH 2n HH nưệp 10 2.1. Facial Recognition SVSf€IS...................... ..- c2: 2. 2111 11211 1111111111111 112 11 1111 11 rệt 10 2.2. Computer Vision Techủn1QU€S.........................- -- 3c 3 31221121128 1231 11 1 1 11 E1 vn rhcry 10 2.2.1. Face [DD€f€CfIOT.....................-- G1 1 211211211211 11 11T TH TH TH TH HH TT HT ng 11 2.2.2. Image Preprocessing .......cecceccesescessesseeseescesceseeseeseeseeseesecsecaecaecaeeaeeeeeeseeseenees 11 2.2.3. Feature EXTAC{IOH....................... LH HT TH TH TH ngàng II 2.3. Gym Management and Member TrackIng.........................- -- --- ¿+ ++s++s£+sx+sx+e+sxeessxs 13 2.4. Related An acc ỨC (11)
  • CHAPTER 3: 050:10)99)09 cam... ẻ...ÔỎ 15 E060) (0)
    • 3.2. Data Preprocessing 7 (16)
      • 3.2.1. Grayscale COMVeTSiON .......ccceccecccessesseesseeceeseeseeeseeesecseeeseeeseenseeeeeteseseeneeeaeees 16 3.2.2. Histogram Equalization 0 (0)
    • 3.3. Model Training n7... e (21)
      • 3.3.1. P.00 0n (0)
      • 3.3.2. LBPH Model “TraIning........................- --- - ¿c2 3111211311511 11 1511111111111 re. 20 3.3.3. Model Evaluation 0ơ °ˆˆễđễ#ễồ".”.€Ầ3 (21)
    • 3.4. User Interface Development ......... ec eceseseeseeeeeeeeeeseececeeseeceeseeseesecseeaeeaeeeeeeeeeeeas 23 1. Member RegISfrafIOT.................... -. c2 3 1111211151 1191 1121111111111 1 111 11111 11x Hư, 24 2. Attendance Tracking 18 (24)
      • 3.4.3. Membership Management ..........ccecceccescescesseseeseeseeeeseceeeeeeceeeeseeseeaeeeeaeeaes 24 3.5. Real-Time Facial Recognition ÍntegTafIOI......................... -- 5s s5 + + s+s£+s£+esexeeeseesses 24 3.6. Data Security and PTIVACY..................... 2c 2c 2. 111111118111 1 11111111111 111 11 11 11 11 1gp 24 (25)
  • CHAPTER 4: SYSTEM IMPLEMENTATIƠN...................... ..SẶcScSxssirirrrrrrrrree 25 4.1. Development EnVITOTIN€TI.....................- + 2 E3 33323 E+E+EE+EE+EEEEEEeEEetEerrrrrrerrerrke 25 4.2. State [Dlapram...................... -. c1 n1 t S991 1 1111 11111111111 TH HT HH TH TH HH 25 4.3. Usecase 0 nan nố (26)
    • 4.4. Sequence Diagram nnn (28)
    • 4.5. Ga n6 ..e (0)
    • 5.2. Training and Validation PhấadÝỶ (35)
    • 5.3. Testing and Recognition PerÍOrmanC©..................... ...--- -- + + + + + *svEskEsereerseesreereerke 35 (0)
      • 5.3.2. False Positives and False NegatIVe€S...................... -.- ác St HH HH t 38 5.4. sài 0i v0. nnỪỪ (39)
      • 5.4.1. Recognition ÁCCUTACV............... ... óc 1211111111211 1111111111111 T1 T1 1 1T ng rệt 40 5.4.2. False Positive Rate (FPR) 07... ............ 4l 5.4.3. False Negative Rate (ENR)........................ Sàn HH HH HH HH nhiệt 4I 5.5. Comparison with Traditional Attendance SŠysfems..................... ..- --- 5555555 *++xs+xs+x 41 5.6. Limitations 000.0 (0)
      • 5.6.2. Variability in Facial EXPT€SSIOTS......................- .- G22. 3221121112 tre 43 5.6.3. Demographic Bias 1 (0)
    • 5.7. CONCIUSION (0)
    • 6.3. Cost-Benefit Analysis 1n... ..Ả (48)
    • 6.4. Future Enhancements and RecommendatIOWS.............................- -¿-- 55+ ++++xs+x>ec+exsxss 47 1. Continuous Model 'TTaIning........................-.. - --- + + xxx rrkee 47 2. Enhanced False Positive Mitigation .........c cc eeceeeeseseseeeceeeeseeseeseeeeeeneentens 48 3. Usability Enhanceim€IfS......................... .-- ¿2+2 3+2 **3**2E+5E£EEEEEEEEEErkrrkrrkrrke 48 6.5. CONCIUSION 1n (48)
  • Chapter 7: ConcẽUSIOTN......................- úc c3 32113211351 3151 1911151111111 1 1111111111 1 1 111 11 T1 HH Hư 49 7.1. Summary of Results 1 (0)
    • 7.2. Effectiveness and Potential for DevelopImen(...........................--- -- 5-5 5+5 s+ss+ss++e+exss 49 1. Time and Effort SaVITE................... .-- 2c 2c 2112612113111 11 1 11 11 11 1111111111 49 (190/8 io a0 (50)
      • 7.2.3. Improved Member Managermer(..........................-- - --- + + + *+e£+t+eserserserserrre 51 7.2.4. Optimized Customer EXP€TI€TIC€......................---- (5 3+ 32333 E*EE+E+EE+eeeeeesrrrerree 51 7.3. Contributions and LIIfAfIOIS.....................- --- c2 3+ 2+ **t**E+EE+EEE+EEsEekrrrkrrkrrrkrrree 51 7.4. Future Development Directions ..0...... cc ecececeeseeseeseeseeeeeeeeeeeeceeceeeeseeseeaeeaeeneeneens 51 7.4.1. Upgrading Recognition Algorithms... cccecesesceeceesceeceeeeseeseeseeaeeeeaeeaees 52 7.4.2. Integration with Attendance and Management Systems (52)
      • 7.4.3. Extension to Other Fields ........................... s6 12x 99 2 919 121 1121 11g ng ưêp 52 (53)

Nội dung

Image Preprocessing Image preprocessing is essential to enhance the quality of facial images before feeding them into the facial recognition system.. Feature extraction is a crucial step

INTRODUCTION 0-5

Problem Statement 00 ccc

In today's fast-paced world, managing gym facilities efficiently and effectively has become a critical task Traditional methods of manual registration and identification of gym members can be time-consuming and prone to errors Additionally, ensuring the security and privacy of gym members' data is of utmost importance.

The problem statement addressed in this thesis is to develop a facial recognition system to manage gym facilities The system aims to provide an automated and accurate solution for identifying gym members, registering their attendance, and maintaining a secure and convenient environment for both gym staff and members.

Problem Solution 8n

To address the challenges in gym management, the proposed solution involves the development of a facial recognition system The system will use state-of-the-art computer vision techniques to recognize and identify gym members based on their facial features By integrating this system into the gym's existing infrastructure, it will streamline the registration process and enhance the overall user experience.

The key components of the problem solution are:

Facial Recognition Model: Develop and train a facial recognition model using OpenCV and LBPH (Local Binary Patterns Histograms) algorithm This model will be capable of recognizing registered gym members' faces accurately.

User Interface: Design and implement a user-friendly interface using Tkinter to facilitate gym staff 1n registering new members and managing member data.

Attendance Tracking: Implement real-time facial recognition to track gym member attendance, providing a seamless and contactless experience.

Data Security: Ensure the protection of sensitive user information and privacy by securely storing member data and incorporating appropriate data protection measures.

Challenges

During the development of the facial recognition system, several challenges need to be addressed:

Accuracy and Robustness: Ensuring high accuracy and robustness of the facial recognition model to correctly identify gym members under various lighting conditions and facial expressions.

Real-time Performance: Implementing the system to perform facial recognition in real- time, enabling swift attendance tracking during peak gym hours.

Data Privacy: Addressing data privacy concerns by adhering to data protection regulations and implementing secure data storage and access mechanisms.

The primary goal of this thesis is to develop a fully functional facial recognition system tailored for gym management The system aims to provide an efficient and secure solution for gym staff to manage member registrations, track attendance, and ensure a smooth gym experience for members.

The study scope encompasses the development, implementation, and evaluation of the facial recognition system It includes:

Collecting and preprocessing facial image data for model training.

Training and optimizing the facial recognition model.

Developing a user interface for gym staff to interact with the system.

Integrating real-time facial recognition for attendance tracking.

Addressing data security and privacy concerns.

The thesis is structured as follows:

Chapter 1: Introduction: This chapter provides an overview of the problem statement, the proposed solution, challenges, goals, and the scope of the study.

Chapter 2: Literature Review: This chapter reviews the existing literature and research related to facial recognition systems, computer vision techniques, and gym management.

Chapter 3: Methodology: In this chapter, the methodology used to develop the facial recognition system will be discussed in detail, including data collection, model training, and system implementation.

Chapter 4: System Implementation: This chapter presents the implementation details of the facial recognition system, including the user interface and real-time facial recognition integration.

Chapter 5: Evaluation and Results: The performance evaluation and results of the facial recognition system will be presented in this chapter.

Chapter 6: Discussion: This chapter discusses the findings, limitations, and potential improvements of the developed facial recognition system.

Chapter 7: Conclusion: The final chapter concludes the thesis, summarizing the achievements and contributions of the research, and suggests future directions for further improvement and application of the facial recognition system in gym management.

References: The reference list of all the sources cited throughout the thesis.

Thesis SÍTUCfUTG - 2 1S nh Tho Thọ TH Tu TH HH HH HH nh TT 8

Facial recognition technology has gained significant attention in recent years due to its wide-ranging applications in various fields, including security, surveillance, and user authentication It involves the use of computer vision algorithms to identify and verify individuals based on their facial features Several facial recognition techniques have been developed, including Eigenfaces, Fisherfaces, and Local Binary Patterns Histograms (LBPH).

The Eigenfaces method, introduced by Turk and Pentland in 1991, uses Principal Component Analysis (PCA) to extract essential facial features from a set of training images. Similarly, Fisherfaces, proposed by Belhumeur et al., is based on Linear Discriminant Analysis (LDA) and aims to maximize the inter-class variations while minimizing intra-class variations.

In recent years, LBPH has emerged as a popular facial recognition technique due to its simplicity and efficiency LBPH encodes facial features by considering local binary patterns in a face image and creates a histogram of these patterns It has shown promising results in various real-world applications and has become the algorithm of choice for many facial recognition systems.

Computer vision techniques play a pivotal role in developing facial recognition systems These techniques enable the system to detect, extract, and process facial features effectively Some fundamental computer vision techniques include:

Computer vision techniques are at the core of developing robust and accurate facial recognition systems These techniques empower the system to detect, extract, and process facial features effectively, laying the foundation for successful facial identification Several fundamental computer vision techniques are utilized in facial recognition systems,

LITERATURE REVIEW LH 2n HH nưệp 10 2.1 Facial Recognition SVSf€IS - c2: 2 2111 11211 1111111111111 112 11 1111 11 rệt 10 2.2 Computer Vision Techủn1QU€S .- 3c 3 31221121128 1231 11 1 1 11 E1 vn rhcry 10 2.2.1 Face [DD€f€CfIOT G1 1 211211211211 11 11T TH TH TH TH HH TT HT ng 11 2.2.2 Image Preprocessing .cecceccesescessesseeseescesceseeseeseeseeseesecsecaecaecaeeaeeeeeeseeseenees 11 2.2.3 Feature EXTAC{IOH LH HT TH TH TH ngàng II 2.3 Gym Management and Member TrackIng .- - ¿+ ++s++s£+sx+sx+e+sxeessxs 13 2.4 Related An acc ỨC

Facial recognition technology has gained significant attention in recent years due to its wide-ranging applications in various fields, including security, surveillance, and user authentication It involves the use of computer vision algorithms to identify and verify individuals based on their facial features Several facial recognition techniques have been developed, including Eigenfaces, Fisherfaces, and Local Binary Patterns Histograms (LBPH).

The Eigenfaces method, introduced by Turk and Pentland in 1991, uses Principal Component Analysis (PCA) to extract essential facial features from a set of training images. Similarly, Fisherfaces, proposed by Belhumeur et al., is based on Linear Discriminant Analysis (LDA) and aims to maximize the inter-class variations while minimizing intra-class variations.

In recent years, LBPH has emerged as a popular facial recognition technique due to its simplicity and efficiency LBPH encodes facial features by considering local binary patterns in a face image and creates a histogram of these patterns It has shown promising results in various real-world applications and has become the algorithm of choice for many facial recognition systems.

Computer vision techniques play a pivotal role in developing facial recognition systems These techniques enable the system to detect, extract, and process facial features effectively Some fundamental computer vision techniques include:

Computer vision techniques are at the core of developing robust and accurate facial recognition systems These techniques empower the system to detect, extract, and process facial features effectively, laying the foundation for successful facial identification Several fundamental computer vision techniques are utilized in facial recognition systems,

Face detection is the process of locating and localizing human faces in an image or video frame Various methods have been proposed for face detection, such as Haar cascades, Histogram of Oriented Gradients (HOG), and Single Shot MultiBox Detector (SSD) Haar cascades, introduced by Viola and Jones, utilize a set of Haar-like features and a trained classifier to detect faces efficiently.

Image preprocessing is essential to enhance the quality of facial images before feeding them into the facial recognition system Common preprocessing techniques include grayscale conversion, histogram equalization, and noise reduction using filters like Gaussian and Bilateral filters These techniques help to improve the robustness and accuracy of the facial recognition process.

Feature extraction aims to identify relevant and distinctive features from facial images that can be used for recognition In addition to the aforementioned PCA and LDA methods, Convolutional Neural Networks (CNNs) have also been extensively employed for feature extraction CNNs have proven to be highly effective in learning hierarchical features and achieving state-of-the-art performance in various computer vision tasks, including facial recognition.

Feature extraction is a crucial step in facial recognition systems, where the goal is to identify relevant and distinctive features from facial images that can be used for recognition purposes Extracting meaningful and discriminative features is essential for accurate and robust identification of individuals in varying conditions.

In addition to the previously mentioned principal component analysis (PCA) and linear discriminant analysis (LDA) methods, Convolutional Neural Networks (CNNs) have gained significant popularity and success in the field of feature extraction for facial recognition. CNNs have demonstrated remarkable capabilities in learning hierarchical features from raw

11 image data, making them highly effective in various computer vision tasks, including facial recognition.

CNNs work by employing multiple layers of convolutional filters, followed by pooling layers, to automatically learn and extract intricate patterns and features from the input images The initial layers detect simple patterns such as edges and corners, while subsequent layers learn more complex and higher-level features, such as facial contours and textures. This hierarchical learning process enables CNNs to capture both local and global features, making them well-suited for facial recognition tasks where facial features can vary in size, shape, and orientation.

The advantage of using CNNs for feature extraction lies in their ability to handle the high dimensionality and variability of facial data efficiently Rather than relying on manual feature engineering, which can be time-consuming and limited in capturing complex patterns, CNNs automatically learn the most discriminative features directly from the data, significantly enhancing the recognition performance.

Moreover, CNNs can be fine-tuned and adapted to specific facial recognition tasks through transfer learning Pretrained CNN models on large-scale datasets, such as ImageNet, can be leveraged as feature extractors for facial recognition systems By fine-tuning these models on a smaller facial recognition dataset, the network can be specialized to detect facial features relevant to the specific task at hand.

In recent years, numerous state-of-the-art facial recognition systems have utilized CNN-based feature extraction methods to achieve unprecedented accuracy and robustness. The combination of deep learning techniques and large-scale facial datasets has propelled the performance of facial recognition systems to new heights, enabling practical applications in diverse real-world scenarios.

2.3 Gym Management and Member Tracking

The management of gym facilities involves various tasks, including member registration, attendance tracking, and membership renewal Traditional gym management systems rely on manual processes, which can be time-consuming and error-prone.

050:10)99)09 cam ẻ ÔỎ 15 E060)

Data Preprocessing 7

Before using the collected facial image data for model training, preprocessing is essential to enhance the data quality and remove any noise or artifacts that may affect the recognition performance.

Image Cleaning and Normalization: In this step, the collected facial images are thoroughly examined to identify and remove any low-quality images or those containing artifacts, such as blurriness, occlusions, or excessive lighting variations Additionally, to reduce the impact of variations in illumination and ensure consistency in the data, the images are normalized to a standard brightness and contrast level.

Facial Region Detection and Alignment: To improve the alignment of facial features across different images, facial region detection techniques are applied These techniques locate key facial landmarks, such as eyes, nose, and mouth, and then align the images based

15 on these landmarks By doing so, the model becomes more robust to variations in head pose and facial expressions.

Image Resizing and Cropping: Consistency in image dimensions is essential for effective model training Therefore, the facial images are resized to a fixed resolution to ensure uniformity across the dataset Additionally, cropping techniques are employed to focus solely on the facial region, excluding any irrelevant background information.

Data Augmentation: To augment the dataset and increase its diversity, data augmentation techniques are applied These techniques involve performing transformations such as rotation, flipping, and scaling on the facial images Data augmentation helps the model generalize better and reduces the risk of overfitting.

Normalization and Standardization: The pixel values of the preprocessed images are normalized and standardized to bring them within a common range Normalization scales the pixel values to fall between 0 and 1, while standardization centers the data around a mean of

0 and a standard deviation of 1 These steps make the model training process more stable and accelerate convergence.

Data Splitting: The preprocessed dataset is divided into three subsets: the training set, validation set, and test set The training set is used to train the facial recognition model, the validation set is used to fine-tune hyperparameters and prevent overfitting, and the test set is reserved for evaluating the final performance of the trained model.

The data preprocessing pipeline will include the following steps:

The facial images will be converted to grayscale to reduce the data dimensionality and computational complexity Grayscale images contain essential facial features while being computationally efficient for subsequent processing.

Grayscale conversion serves two primary purposes:

Dimensionality Reduction: Color images typically have three color channels (red, green, and blue), resulting in higher dimensionality compared to grayscale images, which

16 have only one channel By converting the images to grayscale, we reduce the memory footprint and computational complexity, making it more feasible to process a large number of facial images efficiently.

Preserving Essential Facial Features: Grayscale images retain crucial facial features, such as edges, textures, and facial contours, without the distraction of color information For facial recognition tasks, these essential features play a significant role in identifying individuals and contribute to the robustness of the recognition system.

The grayscale conversion is achieved by taking a weighted sum of the RGB channels for each pixel in the image The formula for grayscale conversion is commonly represented as:

Gray = 0.2989 * R + 0.5870 * G+0.1140 * B where R, G, and B represent the pixel values of the red, green, and blue channels, respectively, and Gray is the resulting grayscale pixel value.

It is important to note that while grayscale conversion reduces dimensionality, it should be performed carefully to ensure that crucial facial information is not lost during the process.

By striking the right balance between data compression and information retention, we can create a dataset that is computationally efficient while retaining the essential facial features necessary for accurate facial recognition.

Following the grayscale conversion, the preprocessed facial images will undergo the subsequent steps in the data preprocessing pipeline, as described in Section 3.2, to prepare them for model training and achieve optimal recognition performance.

Histogram equalization will be applied to improve the contrast and brightness of the facial images This process helps to standardize the intensity levels and enhance the visibility of facial details, making the images more suitable for facial feature extraction.

Original iamge Histogram of original iamge

Image constructed using Equalized Histogram Equalized Histogram ay

The histogram of an image represents the distribution of pixel intensities across its entire range In many cases, facial images may suffer from poor contrast, where the distribution of pixel intensities is concentrated in a narrow range, leading to an image with insufficient brightness variation By applying histogram equalization, the pixel intensities are redistributed across the full range, spreading them out to achieve a more balanced and enhanced contrast.

The result of histogram equalization is that the darker and lighter regions of the facial images are stretched, bringing out finer facial features that might have been hidden in the original low-contrast images This preprocessing step significantly improves the overall image quality, making it easier for subsequent feature extraction methods to identify and extract relevant facial attributes accurately.

Histogram equalization is a simple yet effective technique, particularly useful when dealing with images captured under diverse lighting conditions By standardizing the intensity levels, the facial recognition system becomes more resilient to variations in lighting, thus improving recognition accuracy in challenging environments with varying illumination.

Model Training n7 e

The core component of the facial recognition system is the facial recognition model In this study, the LBPH algorithm will be used due to its simplicity, efficiency, and effectiveness in facial feature encoding.

The dataset will be split into training and testing subsets The training subset will be used to train the model, while the testing subset will be used to evaluate the model's performance and generalization.

The LBPH model will be trained using the training subset of the dataset During training, the model will learn to extract discriminative facial features and encode them as histograms of local binary patterns The model's parameters will be optimized to achieve the best recognition performance. fof ot [eof so [a |

"1 2 oy Lt ofa feat _—ôs PEO |(ị|o

Figure 3.2 Face Recognition using LBPH algorithm

The training process of the LBPH model involves the following steps:

Feature Extraction: For each facial image in the training subset, LBPH extracts local binary patterns from predefined regions (e.g., small square neighborhoods) within the image. These local patterns are encoded as histograms, which serve as the feature representations for the corresponding images.

Labeling: Each facial image in the training subset is associated with a unique label that identifies the individual depicted in the image These labels are used to create ground truth annotations for training the model in a supervised manner.

Model Training: The extracted histograms of local binary patterns, along with their corresponding labels, are used to train the LBPH model The model learns to recognize patterns and features that are discriminative for different individuals in the training data.

Parameter Optimization: During training, the LBPH model's parameters, such as the neighborhood size and the number of points in the local patterns, can be fine-tuned to achieve the best recognition performance Cross-validation techniques may be employed to evaluate the model's performance on validation data and select the optimal hyperparameters.

Model Evaluation: After the model is trained and optimized, its performance is evaluated on the test subset of the dataset The accuracy of facial recognition is measured by comparing the model's predictions with the ground truth labels for the test images.

By following these steps, the LBPH model learns to extract relevant and discriminative facial features, which are crucial for accurate recognition The trained LBPH model can then be used to recognize individuals in real-time facial images, making it a viable choice for facial recognition tasks in scenarios with limited computational resources and smaller datasets.

After model training, the recognition accuracy will be evaluated using the testing subset of the dataset Performance metrics such as accuracy, precision, recall, and Fl-score will be calculated to assess the model's effectiveness in identifying gym members.

Figure 3.3 Comparison between accuracy, precision, recall, and fl score

Once the LBPH model is trained on the training subset of the dataset, it is essential to assess its performance and effectiveness in accurately identifying gym members This evaluation is conducted using the testing subset, which contains facial images that the model has not seen during training Performance metrics are calculated to quantitatively measure the model's recognition accuracy and generalization ability.

The following performance metrics are commonly used to evaluate the facial recognition system:

Accuracy: Accuracy represents the proportion of correctly classified samples out of the total number of samples in the testing subset It is a fundamental metric to measure the overall performance of the model.

Precision: Precision is the ratio of true positive predictions to the total number of positive predictions It measures the model's ability to correctly identify positive samples (correctly recognized gym members) and avoid false positives (incorrectly recognized non- members).

Recall (Sensitivity): Recall, also known as sensitivity, is the ratio of true positive predictions to the total number of actual positive samples in the dataset It quantifies the model's ability to correctly detect all the positive samples in the testing subset.

Fl-score: The Fl-score is the harmonic mean of precision and recall It provides a balanced measure that takes both precision and recall into account, making it useful when classes are imbalanced.

User Interface Development ec eceseseeseeeeeeeeeeseececeeseeceeseeseesecseeaeeaeeeeeeeeeeeas 23 1 Member RegISfrafIOT - c2 3 1111211151 1191 1121111111111 1 111 11111 11x Hư, 24 2 Attendance Tracking 18

To facilitate gym staff in interacting with the facial recognition system, a user-friendly interface will be developed using Tkinter, a Python GUI library.

Gym staff will be able to register new members by capturing their facial images and entering their relevant information, such as name, age, and contact details.

The user interface will provide a real-time attendance tracking feature that enables gym staff to record the entry and exit times of gym members.

The user interface will also allow gym staff to manage member data, update member information, and track membership status.

3.5 Real-Time Facial Recognition Integration

For real-time attendance tracking, the facial recognition model will be integrated with the gym's camera system As gym members enter the facility, their facial images will be captured by the camera, and the recognition model will identify and verify their identity The attendance records will be automatically updated in the system.

Data security and privacy are of utmost importance in a facial recognition system To protect sensitive member information, measures will be implemented to ensure secure data storage and access Encryption techniques will be employed to safeguard member data, and access control mechanisms will be put in place to restrict unauthorized access.

SYSTEM IMPLEMENTATIƠN SẶcScSxssirirrrrrrrrree 25 4.1 Development EnVITOTIN€TI - + 2 E3 33323 E+E+EE+EE+EEEEEEeEEetEerrrrrrerrerrke 25 4.2 State [Dlapram - c1 n1 t S991 1 1111 11111111111 TH HT HH TH TH HH 25 4.3 Usecase 0 nan nố

Sequence Diagram nnn

The Gymer scan their face at the entrance, if the System Records recognize their face, it continue ‘Record Attendance’ and then it notice the Gymer by displaying customer information on screen.

Face Recognition System Records Aliendonce Visitor Records

Camera already turned on by Manager when he/she enter the gym When the customer comes in, the camera scan their face By using its AI setted up before, it brings out two questions: “Does this face exists in the database?’, ‘Is it a customer?’ If the answer all yes, the screen display the customer information by the data of that customer which was saved before If the answer is no, it display nothing beside that guest on screen (not their information).

FACE RECOGNITION IN GYM MANAGEMENT SYSTEM

Does this face exists in the database?

Is it a customer? isplay Customer Name”

The user interface (UI) will be designed to provide a seamless and intuitive experience for gym staff in managing member registrations and attendance tracking The UI will consist of various screens with distinct functionalities.

FACE RECOGNITION FOR GYM MANAGEMENT SYSTEM

The member registration screen will allow gym staff to add new members to the system It will include fields for entering member details such as name, age, gender, and contact information Gym staff will also be able to capture the member's facial image using the integrated camera.

The attendance tracking screen will display a real-time feed from the camera, detecting and recognizing faces as gym members enter the facility The system will automatically mark the entry time when a recognized member is detected.

The membership management screen will provide an overview of all registered gym members Gym staff can update member information, renew memberships, and view membership status.

Figure 4.8 History saves in Excel

The facial recognition model developed in Chapter 3 will be integrated into the system for real-time recognition As a gym member approaches the entrance, the camera will capture their facial image, and the model will identify and verify their identity based on the stored facial features If the member is recognized, their attendance record will be updated with the current timestamp.

To store member information and attendance records, data serialization using the Pickle library will be employed The system will create a 'users.pickle' file to store member details and associated facial features This file will be updated whenever new members are registered or existing member information is modified.

Additionally, the system will maintain an 'attendance.log' file to record attendance entries and exits Each entry in the log file will include the member's name, entry time, and exit time if applicable.

To ensure real-time performance during facial recognition, various optimization techniques will be applied These include:

Face detection region-of-interest (ROI) filtering: The system will focus on specific regions around the entrance area, minimizing the processing load and improving response time.

Multi-threading: The recognition process will run on a separate thread to avoid blocking the user interface and provide a seamless experience.

Caching: The system will cache facial features to reduce redundant calculations during recognition.

Once the system implementation is complete, extensive testing and validation will be performed to assess its accuracy, robustness, and usability The testing will involve scenarios with different lighting conditions, facial expressions, and variations in member appearances. The performance metrics, such as accuracy, recognition speed, and memory consumption, will be recorded and analyzed.

Before deploying the system in the gym, it will be integrated with the gym's existing infrastructure The camera system will be installed at the entrance, and the user interface will be deployed on gym staff computers The system will be connected to the gym database to synchronize member information and attendance records.

FACE RECOGNITION FOR GYM MANAGEMENT SYSTEM

Figure 4.9 Camera scans and shows customer’s information on screen

The evaluation and results chapter focuses on the dataset used for training, testing, and validating the facial recognition system The dataset played a crucial role in assessing the system's accuracy, robustness, and performance.

The dataset consisted of a diverse collection of facial images, comprising both gym members and non-members Each member's image was captured from different angles and under various lighting conditions to simulate real-world scenarios The dataset also included images with different facial expressions, glasses, and facial hair to introduce variations in member appearances.

To ensure fairness and inclusivity, the non-member images were carefully selected to represent a broad demographic, encompassing various age groups, ethnicities, and genders. The inclusion of non-members' images aimed to test the system's ability to differentiate between registered members and non-members accurately.

The facial recognition system was trained using a supervised learning approach with the labeled dataset The face detection algorithm localized facial regions in each image, and the facial landmarks were identified to align the facial images consistently.

The training data comprised facial images as input and their corresponding labels as output Each label represented a unique identifier for each registered gym member The LBPH (Local Binary Patterns Histograms) algorithm was employed to extract facial features and create a compact facial feature representation for each member.

Training and Validation PhấadÝỶ

The facial recognition system was trained using a supervised learning approach with the labeled dataset The face detection algorithm localized facial regions in each image, and the facial landmarks were identified to align the facial images consistently.

The training data comprised facial images as input and their corresponding labels as output Each label represented a unique identifier for each registered gym member The LBPH (Local Binary Patterns Histograms) algorithm was employed to extract facial features and create a compact facial feature representation for each member.

To validate the trained model's performance, a portion of the dataset was reserved for validation The validation dataset was used to assess the model's accuracy and identify potential overfitting issues The hyperparameters of the model, such as the threshold for confidence in recognition, were tuned based on the validation results.

After training and validation, the facial recognition model was tested using a separate test dataset, which was distinct from both the training and validation datasets The test dataset comprised facial images captured from the gym's entrance during the system's deployment phase.

Following the training and validation phases, the facial recognition model was put to the ultimate test using a distinct and independent test dataset The test dataset was carefully curated to ensure that it did not overlap with either the training or validation datasets, thus providing an unbiased evaluation of the model's real-world recognition performance.

The test dataset consisted of facial images captured from the gym's entrance during the system's deployment phase These images represented real-world scenarios, where gym members were arriving at the gym and being recognized by the facial recognition system in real-time The diversity of lighting conditions, facial expressions, and head poses in the test dataset aimed to challenge the model's ability to generalize to new and previously unseen data.

The recognition performance of the facial recognition model was measured using the evaluation metrics discussed in Section 3.3.3 These metrics, including accuracy, precision, recall, and Fl-score, provided quantitative insights into the model's effectiveness in correctly identifying gym members and its robustness against potential false positives or false negatives.

Additionally, a confusion matrix was constructed based on the test results The confusion matrix allowed a detailed analysis of the model's predictions, displaying true positive, true negative, false positive, and false negative counts This visual representation of the model's performance offered a deeper understanding of the types of errors made by the system.

The testing phase was crucial in assessing the real-world viability of the developed facial recognition system for gym member identification By evaluating the system's

35 performance on unseen data, we could determine its practical applicability and identify any potential challenges or limitations.

To ensure the reliability of the testing results, measures were taken to prevent data leakage and maintain data separation between training, validation, and test datasets Careful attention was given to data preprocessing, model training, and hyperparameter tuning to avoid overfitting and improve generalization capabilities.

The testing and recognition performance results provided valuable feedback on the model's strengths and weaknesses, enabling potential refinements to enhance its accuracy and overall performance in the gym management context The insights gained from this evaluation phase played a crucial role in validating the effectiveness of the facial recognition system in real-world settings and guiding any necessary improvements or adjustments before full-scale deployment in the gym facility.

The testing phase aimed to evaluate the recognition accuracy of the system in real- world conditions The system was assessed on its ability to correctly identify registered gym members and differentiate them from non-members.

During the testing phase, the system processed a diverse set of facial images captured at the gym's entrance, which represented real-world scenarios of gym members arriving for their workouts These images were acquired under varying lighting conditions, facial expressions, and head poses, mimicking the challenges encountered in practical gym management settings.

To measure recognition accuracy, the system compared the facial features extracted from the test images against the features stored in the database of registered gym members.

If the system correctly matched a test image with the corresponding identity in the database, it was deemed a true positive recognition Conversely, if the system identified a test image as a registered gym member incorrectly, or if it failed to identify a gym member from the test images, it was categorized as a false positive or false negative recognition, respectively.

The recognition accuracy was then calculated as the percentage of correctly recognized gym members out of the total number of test images Additionally, precision, recall, and the Fl-score were computed to provide a more comprehensive evaluation of the system's performance.

Precision represents the proportion of correctly identified gym members (true positives) out of all the positive predictions (both true positives and false positives) It measures the system's ability to avoid misclassifying non-members as gym members.

Testing and Recognition PerÍOrmanC© - + + + + + *svEskEsereerseesreereerke 35

5.3.2 False Positives and False Negatives

As with any facial recognition system, false positives and false negatives were encountered during testing False positives occurred when the system incorrectly identified a non-member as a registered gym member False negatives occurred when the system failed to recognize a registered member.

During the testing phase of the facial recognition system, as with any recognition technology, false positives and false negatives were encountered These two types of errors are common challenges in facial recognition systems and play a critical role in assessing the system's accuracy and reliability.

False positives occurred when the facial recognition system incorrectly identified a non-member as a registered gym member In other words, the system incorrectly matched a test image to a member's identity in the database when the individual was not a registered gym member False positives can be influenced by various factors, such as lighting conditions, facial expressions, and similarities between individuals' facial features These errors can be particularly concerning in gym management, as they may grant unauthorized access to non-members.

On the other hand, false negatives occurred when the system failed to recognize a registered gym member In these cases, the system could not correctly match a test image to the corresponding identity in the database, resulting in a failure to identify a gym member who should have been recognized False negatives can arise due to variations in appearance, such as changes in facial expressions, head poses, or different hairstyles, which may affect the system's ability to align and match facial features accurately.

Addressing False Positives and False Negatives:

Reducing false positives and false negatives is essential to improve the overall accuracy and reliability of the facial recognition system Several strategies can be employed to mitigate these errors:

Threshold Adjustment: The system's decision threshold, which determines the similarity threshold for a positive match, can be fine-tuned By adjusting this threshold, the balance between false positives and false negatives can be optimized, depending on the specific requirements of the gym management scenario.

Data Augmentation: Increasing the diversity of the training data through data augmentation techniques can enhance the system's robustness and generalization capabilities. Augmented data can represent a broader range of facial variations, reducing the likelihood of false negatives caused by different poses or expressions.

Anti-Spoofing Measures: Implementing robust face anti-spoofing techniques can help minimize false positives caused by presentation attacks, such as using photographs or videos to deceive the system.

Model Fine-Tuning: Fine-tuning the facial recognition model using additional data, including images of individuals who exhibit similar facial features to the false positive cases, can help improve recognition accuracy.

Feedback and Continuous Improvement: Gathering feedback from users and regularly updating the system based on user experiences and encounters with false positives and false negatives can lead to ongoing improvements and optimizations.

Understanding the occurrence of false positives and false negatives provides valuable insights into the system's strengths and areas for improvement By systematically addressing these errors and implementing strategies to minimize their occurrence, the facial recognition system can achieve higher accuracy and reliability, making it a more valuable asset for gym management and member identification.

The occurrence of false positives and false negatives was relatively low and was analyzed to identify potential causes In some instances, false positives were attributed to

39 similarities in facial features between non-members and registered members False negatives were mainly due to unfavorable lighting conditions or occlusions on facial images.

Several performance metrics were calculated to quantify the facial recognition system's performance The metrics included:

The recognition accuracy metric represented the percentage of correctly recognized gym members out of the total number of tested images.

To calculate recognition accuracy, the system compared the facial features extracted from the test images against the features stored in the database of registered gym members.

If the system correctly matched a test image with the corresponding identity in the database, it was considered a true positive recognition Conversely, if the system identified a test image as a registered gym member incorrectly, or if it failed to identify a gym member from the test images, it was categorized as a false positive or false negative recognition, respectively.

Recognition accuracy is a fundamental performance metric that reflects the system's ability to correctly identify gym members in real-world scenarios A higher recognition accuracy indicates a more reliable and effective facial recognition system Conversely, a lower accuracy may highlight potential issues that need to be addressed, such as challenges in handling varying lighting conditions, facial expressions, or head poses.

By evaluating recognition accuracy, gym management can gauge the system's performance in accurately identifying registered members and ensuring secure access control A high recognition accuracy provides confidence in the system's ability to operate efficiently, reducing the risk of unauthorized entry and streamlining the attendance tracking process.

However, it's essential to consider the system's recognition accuracy in conjunction with other performance metrics, such as precision, recall, and the Fl-score, for a comprehensive evaluation While recognition accuracy provides an overall view of the

40 system's performance, these additional metrics offer insights into specific aspects, such as false positives and false negatives, which may be critical for system optimization and fine- tuning.

Moreover, recognition accuracy can be influenced by factors such as the size and diversity of the training dataset, the quality of the facial images, and the effectiveness of the feature extraction and matching algorithms As such, continuous monitoring and periodic updates to the system, along with data augmentation and algorithm refinement, can contribute to improving recognition accuracy over time.

The false positive rate measured the percentage of non-members incorrectly identified as registered gym members.

The false negative rate represented the percentage of registered gym members incorrectly not recognized by the system.

5.5 Comparison with Traditional Attendance Systems

To assess the facial recognition system's efficiency and effectiveness, a comparison was made with traditional attendance tracking methods, such as manual entry using ID cards or sign-in sheets.

The results revealed that the facial recognition system significantly reduced the time and effort required for attendance tracking The automated nature of the system minimized the need for manual data entry and streamlined the check-in process for gym members.

Additionally, the facial recognition system provided real-time attendance updates, ensuring accurate and up-to-date attendance records compared to potential delays or inaccuracies in manual systems.

Despite the facial recognition system's successes, certain limitations and challenges were encountered during its evaluation and deployment.

Some of the key limitations and challenges include:

Cost-Benefit Analysis 1n Ả

The cost-benefit analysis examined the financial implications of implementing the facial recognition system compared to the benefits it provided The initial setup cost, ongoing maintenance expenses, and staff training costs were considered against the potential savings in administrative labor and time due to automation.

The analysis revealed that while the initial setup cost and integration efforts were substantial, the long-term benefits, including improved security, streamlined attendance tracking, and enhanced member management, outweighed the initial investment The system's efficiency in reducing administrative overhead and enhancing member experience justified its implementation from a cost-benefit perspective.

Future Enhancements and RecommendatIOWS - -¿ 55+ ++++xs+x>ec+exsxss 47 1 Continuous Model 'TTaIning - - - + + xxx rrkee 47 2 Enhanced False Positive Mitigation c cc eeceeeeseseseeeceeeeseeseeseeeeeeneentens 48 3 Usability Enhanceim€IfS . ¿2+2 3+2 **3**2E+5E£EEEEEEEEEErkrrkrrkrrke 48 6.5 CONCIUSION 1n

The discussion chapter also addresses potential areas for improvement and future enhancements to the facial recognition system Based on the evaluation results and user feedback, several recommendations were proposed to enhance the system's performance and user experience.

To further improve recognition accuracy, continuous model training with updated member data is recommended Regularly retraining the facial recognition model using new

41 facial images from registered members will help adapt the model to variations in appearance and improve its performance over time.

Efforts to minimize false positives, especially in cases of facial feature similarities among non-members, could be explored Leveraging additional biometric identification methods, such as fingerprint recognition, in combination with facial recognition may provide a more robust and reliable identification process.

User interface enhancements could be implemented based on user feedback Improving the user interface's intuitiveness and providing additional features, such as search and filter options for member management, could further enhance user experience.

The discussion chapter concludes by summarizing the facial recognition system's performance, addressing the key findings, and highlighting its impact on gym operations. The system's contribution to streamlined attendance tracking, enhanced security, and efficient member management is emphasized.

The chapter also acknowledges the system's limitations and ethical considerations, emphasizing the importance of data privacy and transparency in implementing facial recognition technology.

ConcẽUSIOTN - úc c3 32113211351 3151 1911151111111 1 1111111111 1 1 111 11 T1 HH Hư 49 7.1 Summary of Results 1

Effectiveness and Potential for DevelopImen( - 5-5 5+5 s+ss+ss++e+exss 49 1 Time and Effort SaVITE 2c 2c 2112612113111 11 1 11 11 11 1111111111 49 (190/8 io a0

The facial recognition system has proven to be an effective and promising technology for gym management and sports facilities Its quick and accurate facial identification brings several advantages, including:

The system minimizes the time and effort required for attendance, allowing staff to focus on other important tasks.

One of the significant benefits of implementing the facial recognition system in gym management is its ability to minimize the time and effort required for attendance tracking. Traditionally, manual attendance tracking methods, such as using physical cards or sign-in sheets, can be time-consuming and prone to errors The facial recognition system streamlines this process, offering a fast and efficient solution for member identification.

With the facial recognition system in place, gym staff no longer need to manually check and verify each member's identity during entry As members approach the gym entrance, the system automatically captures and matches their facial images against the database of registered members This real-time recognition process is swift and seamless, allowing members to enter the facility with minimal delays.

The time and effort saved by the facial recognition system are significant, especially during peak hours when gym foot traffic is high Staff can focus on other important tasks, such as providing customer support, assisting new members, or ensuring the gym's smooth operation By automating the attendance tracking process, the system empowers gym staff to allocate their time and efforts more efficiently, leading to improved productivity and member satisfaction.

Moreover, the facial recognition system reduces the administrative burden associated with manual attendance records With attendance data automatically recorded and updated in the system, the need for manual data entry and record-keeping is eliminated This not only saves time but also reduces the likelihood of human errors, ensuring more accurate and reliable attendance records.

The time and effort saved through the facial recognition system contribute to a more streamlined and efficient gym management process Gym staff can focus on creating a welcoming and engaging environment for members, leading to a better overall member experience Additionally, the increased efficiency in attendance tracking can positively impact member retention, as members appreciate the convenience and smooth entry process provided by the system.

By optimizing time and effort allocation, the facial recognition system enhances the operational efficiency of gym management As a result, staff can concentrate on delivering exceptional customer service and personalized assistance, strengthening the gym's reputation and fostering a sense of community among members.

Facial recognition ensures precise and unique identification of each member, increasing security and preventing fraud or intrusion attempts.

The system helps enhance member information management, including joining time, expiration dates, and providing accurate activity history for each member.

Members are served quickly and efficiently, improving customer experience, satisfaction, and loyalty.

Although many positive outcomes have been achieved, the facial recognition system also faces certain challenges and limitations, as discussed in Chapter 5 Continual research and improvement are essential to optimize the system's efficiency and reliability.

The project "Facial Recognition System for Gym Management" contributes to the application of facial recognition technology in the domain of sports facilities and gym management The developed and implemented system brings significant benefits in terms of time and performance optimization in attendance and member management.

However, the system implementation requires technical factors, hardware integration, and adequate preparation from gym facilities For larger facilities with complex attendance processes, system integration and expansion may demand considerable time and resources.

The project marks the initial step in applying facial recognition technology to gym management and sports facilities In the future, the system can be expanded and improved with more advanced features, including:

Research and develop more advanced facial recognition algorithms to enhance the system's accuracy and stability.

7.4.2 Integration with Attendance and Management Systems

Optimize the integration of the attendance system with other management systems, such as member management, scheduling, and other gym services.

Explore the potential use of facial recognition technology in other domains, such as security, surveillance, and customer identification.

The project "Facial Recognition System for Gym Management" has successfully built an efficient and promising system for managing and tracking gym members Its performance and applicability have been evaluated, and potential future development directions have been proposed The implementation of facial recognition technology in sports facilities and gym management continues to play a significant role in optimizing the attendance process and improving member management efficiency.

Ngày đăng: 02/10/2024, 05:20

TRÍCH ĐOẠN

TÀI LIỆU CÙNG NGƯỜI DÙNG

TÀI LIỆU LIÊN QUAN

w