Improvement of Face and Eye Detection Performance by Using Multi-task Cascaded Convolutional Networks

Robin, Mahmudul Hasan; Rahman, Md. Minhaz Ur; Taief, Abu Mohammad; Eity, Qamrun Nahar

doi:10.1109/tensymp50017.2020.9230756

Cited by 21 publications

(2 citation statements)

References 9 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…The input device of this system was an OpenMV camera. The Open MV camera took a facial image and then detected the eye part of the motorist (Robin, 2020). Figure 3 is a driver's face detection.…”

Section: Methodsmentioning

confidence: 99%

Motorized Driving Safety System Using Eye Detection Analysis Method

Sumardi

Sudarti

Muldayani

et al. 2022

jppipa, pendidikan ipa, fisika, biologi, kimia

View full text Add to dashboard Cite

Traffic accident, particularly two-wheeled vehicles, is a problem for the Government, especially the Resort Police Traffic Unit (Satlantas Polres). The factor that causes the traffic accident incident is divided into three types, namely human factor, vehicle factor, and road or environment factor. The human factor is the most common factor for an accident. Fatigue factor that causes someone to feel sleepy while driving often results in a traffic accident. Based on the problem, the researcher wanted to create a technology innovation of a motorized driving safety system in the form of a helmet. The researcher made an innovation of a helmet that can detect drowsiness through the driver's eye blink duration. The drowsiness will be detected by using a camera sensor. The camera sensor used was Open MV camera. The method used in detecting sleepy drivers was the eye detection analysis method. The method enable detection based on the data of the duration of eye condition when it is closed and open. The closed eye has a low RBG mean value of 110-113 and an RBG median value of 99-109. Whereas opened eye has a higher RGB mean value of 179-206 and RGB median value of 178-206. The result of the research showed that someone's sleepy condition occurred when closing their eyes for more than 0.4 seconds to 4 seconds. The helmet is also equipped with GPS to monitor the position in the event of an accident as an emergency response effort.

show abstract

Section: Methodsmentioning

confidence: 99%

Motorized Driving Safety System Using Eye Detection Analysis Method

Sumardi

Sudarti

Muldayani

et al. 2022

jppipa, pendidikan ipa, fisika, biologi, kimia

View full text Add to dashboard Cite

show abstract

“…The local dataset considering of data having 10 virtual and 14 EEG raw data was used to test the model (10) . The enhanced face and eye detection technique using Cascaded multi-task convolutional network is performed (11) . The MTCNN and Haar-based Cascade classifier is used to detect the face and eyes.…”

Section: Introductionmentioning

confidence: 99%

Performance Evaluation of Fusion Based Efficient Algorithm for Facial Expression Recognition

Harakannanavar¹,

Sapnakumari²,

Ramachandra³

et al. 2023

IJST

View full text Add to dashboard Cite

Objectives: To develop face expression recognition system using JAFFE database and to evaluate the performance of the face expression recognition models. Methods: This study used the FER model based on modified-HoG (Histogram of oriented gradient), LBP (Local Binary Patterns) and Fast Key point detector and BRIEF descriptor (FKBD) to extract the significant features of JAFFE dataset. The features extracted using HoG, LBP and FKBD techniques form a feature vector. Then, the fusion of all the features is carried out at the feature level. The multiclass SVM and KNN classifiers are used to recognize the facial expressions, effectively. Findings: In this work, an effort is made to develop a robust FER model using JAFFE database. It is recorded that, based on the experimental results, the proposed model suits better with a performance rate of 98.26% for SVM and 96.51% for KNN, when compared with the different state-of-the-art methods. Novelty: Many FER models have been developed and adopted for enhancing their quality and to extract the facial features using transform and frequency domains. It is observed that, maximum approaches are based on generating the texture features. The fusion at the feature level using modified HoG, LBP and FKBD is performed and the SVM model is more compatible when compared with other classifiers and it supports one-to-one and one-to-many comparisons' technique.

show abstract