Multi-Pose Face Recognition Based on Deep Learning in Unconstrained Scene

Ruan, Shuai; Tang, Chaowei; Zhou, Xu; Jin, Zhuoyi; Chen, Shiyu; Wen, Haotian; Liu, Hongbin; Tang, Dong Sheng

doi:10.3390/app10134669

Cited by 9 publications

(4 citation statements)

References 28 publications

(40 reference statements)

Supporting

Mentioning

Contrasting

Order By: Relevance

“…The existing computational intelligence learning models are unable to accurately distinguish faces in images with varying perspectives [17]. This is because changes in surface and pattern caused by shifting perspectives often outweigh the differences between individuals, as highlighted in [18], [19]. The more recently developed face recognition techniques for face recognition can be broadly categorized into two groups [20].…”

Section: A Depth Pose Alignment Of Imagementioning

confidence: 99%

Deep Learning Based Multi Pose Human Face Matching System

Sohail,

Shoukat,

Khan

et al. 2024

IEEE Access

View full text Add to dashboard Cite

Current techniques for multi-pose human face matching yield suboptimal outcomes because of the intricate nature of pose equalization and face rotation. Deep learning models, such as YOLO-V5, etc., that have been proposed to tackle these complexities, suffer from slow frame matching speeds and therefore exhibit low face recognition accuracy. Concerning this, certain literature investigated multi-pose human face detection systems; however, those studies are of elementary level and do not adequately analyze the utility of those systems. To fill this research gap, we propose a real-time face matching algorithm based on YOLO-V5. Our algorithm utilizes multi-pose human patterns and considers various face orientations, including organizational faces and left, right, top, and bottom alignments, to recognize multiple aspects of people. Using face poses, the algorithm identifies face positions in a dataset of images obtained from mixed pattern live streams, and compares faces with a specific piece of the face that has a relatively similar spectrum for matching with a given dataset. Once a match is found, the algorithm displays the face on Google Colab, collected during the learning phase with the Robo-flow key, and tracks it using the YOLO-V5 face monitor. Alignment variations are broken up into different positions, where each type of face is uniquely learned to have its own study demonstrated. This method offers several benefits for identifying and monitoring humans using their labeling tag as a pattern name, including high face-matching accuracy and minimum speed of owing face-to-pose variations. Furthermore, the algorithm addresses the face rotation issue by introducing a mixture of error functions for execution time, accuracy loss, frame-wise failure, and identity loss, attempting to guide the authenticity of the produced image frame. Experimental results confirm effectiveness of the algorithm in terms of improved accuracy and reduced delay in the face-matching paradigm.

show abstract

Section: A Depth Pose Alignment Of Imagementioning

confidence: 99%

Deep Learning Based Multi Pose Human Face Matching System

Sohail,

Shoukat,

Khan

et al. 2024

IEEE Access

View full text Add to dashboard Cite

show abstract

“…The tire normal force and road roughness in the LSTM model was also estimated through training, verification, and testing processes. In this study, shuffle division cross-validation was used because it allows the alleviation of overfitting concerns and rapid operations for large data sets [29]. The hyperparameters used for training were maximum epochs, initial learning rate, learning rate drop period, learning rate drop factor, and mini-batch size.…”

Section: B Training Data Acquisitionmentioning

confidence: 99%

Simultaneous Estimation of Unknown Road Roughness Input and Tire Normal Forces Based on a Long Short-Term Memory Model

Kim

2022

IEEE Access

View full text Add to dashboard Cite

This paper reports an initial study on the simultaneous estimation of unknown road roughness input and tire normal forces for automotive vehicles using a long short-term memory (LSTM) model. Active safety systems and the improvement of ride comfort using vehicle information have garnered increasing attention in the automotive industry. In particular, active safety systems rely significantly on road roughness data and the normal force of the tires. If these factors can be measured in real-time for a driving vehicle, the measured data can be used for automotive control systems for tasks, such as semi-active and active suspension control, rolling motion control, and torque vectoring. However, it is typically difficult to measure the road roughness and tire normal force directly in real-time by mounting physical sensors on the vehicle. In this study, we explore the simultaneous estimation of these factors using an LSTM model that requires only timeseries data of the vehicle body. The LSTM model is implemented by using MATLAB/Simulink and includes data preprocessing, learning, and verification steps. To evaluate the estimation performance of the LSTM model, we compared it with a Kalman filter and used CarSim vehicle simulation software to simulate and interpret the dynamic behavior of vehicles. INDEX TERMSlong short-term memory model; discrete Kalman filter-unknown input; 7-DOF full-car suspension model; tire normal forces; road roughness SUNG JIN IM received his B.S. degree in Mechanical Engineering from

show abstract

“…Renowned algorithms within this methodology encompass R-FCN [1], Fast RCNN [2], and Mask-RCNN [3]. For instance, a face detection system based on R-FCN was proposed by Ruan et al [4]. Despite achieving commendable accuracy in intricate backgrounds, its detection latency was found to be extensive.…”

Section: Introductionmentioning

confidence: 99%

Enhanced Campus Security Target Detection Using a Refined YOLOv7 Approach

Cao,

2023

View full text Add to dashboard Cite

In many educational institutions, safety management traditionally depends upon manual video surveillance, leading to potential delays in the identification and alerting of perilous activities, notably the possession of controlled knives and smoking behaviors exhibited by students. These activities possess significant consequences for both the psychological and physical well-being of students. Recognizing this pressing need, an augmented object detection method for campus security, rooted in YOLOv7, is presented. The EIoU (Efficient Intersection over Union) loss function has been substituted to expedite model convergence and heighten detection fidelity. Additionally, the integration of the CBAM (Convolutional Block Attention Module) attention mechanism with the DCNv2 (Deformable ConvNets v2) deformable convolutional kernel not only mitigates the challenge of information inundation but also enhances feature extraction capabilities, facilitating adjustments to geometric deformations. Experimental findings indicate that this proposed method achieves a detection accuracy of 92.6% across various categories on a dataset comprising three categories, spanning a total of 4500 images, and attains an mAP of 96.4%. In comparison to the conventional YOLOv7 algorithm, enhancements in detection accuracy and mAP by 6.9% and 6.6%, respectively, have been observed, affirming the efficacy of the presented algorithm.

show abstract

Multi-Pose Face Recognition Based on Deep Learning in Unconstrained Scene

Cited by 9 publications

References 28 publications

Deep Learning Based Multi Pose Human Face Matching System

Deep Learning Based Multi Pose Human Face Matching System

Simultaneous Estimation of Unknown Road Roughness Input and Tire Normal Forces Based on a Long Short-Term Memory Model

Enhanced Campus Security Target Detection Using a Refined YOLOv7 Approach

Contact Info

Product

Resources

About