scholarly journals Cost-Effective CNNs for Real-Time Micro-Expression Recognition

2020 ◽  
Vol 10 (14) ◽  
pp. 4959
Author(s):  
Reda Belaiche ◽  
Yu Liu ◽  
Cyrille Migniot ◽  
Dominique Ginhac ◽  
Fan Yang
Keyword(s):  
Machine Learning ◽  
Computer Vision ◽  
Real Time ◽  
Video Clip ◽  
Cost Effective ◽  
Compact Representation ◽  
Expression Recognition ◽  
Memory Space ◽  
Computational Resources ◽  
Micro Expression

Micro-Expression (ME) recognition is a hot topic in computer vision as it presents a gateway to capture and understand daily human emotions. It is nonetheless a challenging problem due to ME typically being transient (lasting less than 200 ms) and subtle. Recent advances in machine learning enable new and effective methods to be adopted for solving diverse computer vision tasks. In particular, the use of deep learning techniques on large datasets outperforms classical approaches based on classical machine learning which rely on hand-crafted features. Even though available datasets for spontaneous ME are scarce and much smaller, using off-the-shelf Convolutional Neural Networks (CNNs) still demonstrates satisfactory classification results. However, these networks are intense in terms of memory consumption and computational resources. This poses great challenges when deploying CNN-based solutions in many applications, such as driver monitoring and comprehension recognition in virtual classrooms, which demand fast and accurate recognition. As these networks were initially designed for tasks of different domains, they are over-parameterized and need to be optimized for ME recognition. In this paper, we propose a new network based on the well-known ResNet18 which we optimized for ME classification in two ways. Firstly, we reduced the depth of the network by removing residual layers. Secondly, we introduced a more compact representation of optical flow used as input to the network. We present extensive experiments and demonstrate that the proposed network obtains accuracies comparable to the state-of-the-art methods while significantly reducing the necessary memory space. Our best classification accuracy was 60.17% on the challenging composite dataset containing five objectives classes. Our method takes only 24.6 ms for classifying a ME video clip (less than the occurrence time of the shortest ME which lasts 40 ms). Our CNN design is suitable for real-time embedded applications with limited memory and computing resources.

Deep Convolutional Neural Network with Optical Flow for Facial Micro-Expression Recognition

2019 ◽  
Vol 29 (01) ◽  
pp. 2050006 ◽  
Author(s):  
Qiuyu Li ◽  
Jun Yu ◽  
Toru Kurihara ◽  
Haiyan Zhang ◽  
Shu Zhan
Keyword(s):  
Neural Network ◽  
Optical Flow ◽  
Video Clip ◽  
Support Vector ◽  
Expression Recognition ◽  
Convolutional Network ◽  
Facial Landmarks ◽  
Flow Features ◽  
Optical Flow Features ◽  
Micro Expression

Micro-expression is a kind of brief facial movements which could not be controlled by the nervous system. Micro-expression indicates that a person is hiding his true emotion consciously. Micro-expression recognition has various potential applications in public security and clinical medicine. Researches are focused on the automatic micro-expression recognition, because it is hard to recognize the micro-expression by people themselves. This research proposed a novel algorithm for automatic micro-expression recognition which combined a deep multi-task convolutional network for detecting the facial landmarks and a fused deep convolutional network for estimating the optical flow features of the micro-expression. First, the deep multi-task convolutional network is employed to detect facial landmarks with the manifold-related tasks for dividing the facial region. Furthermore, a fused convolutional network is applied for extracting the optical flow features from the facial regions which contain the muscle changes when the micro-expression appears. Because each video clip has many frames, the original optical flow features of the whole video clip will have high number of dimensions and redundant information. This research revises the optical flow features for reducing the redundant dimensions. Finally, a revised optical flow feature is applied for refining the information of the features and a support vector machine classifier is adopted for recognizing the micro-expression. The main contribution of work is combining the deep multi-task learning neural network and the fusion optical flow network for micro-expression recognition and revising the optical flow features for reducing the redundant dimensions. The results of experiments on two spontaneous micro-expression databases prove that our method achieved competitive performance in micro-expression recognition.

scholarly journals Real-Time Evaluation of Breast Self-Examination Using Computer Vision

2014 ◽  
Vol 2014 ◽  
pp. 1-12 ◽  
Author(s):  
Eman Mohammadi ◽  
Elmer P. Dadios ◽  
Laurence A. Gan Lim ◽  
Melvin K. Cabatuan ◽  
Raouf N. G. Naguib ◽  
...  
Keyword(s):  
Breast Cancer ◽  
Computer Vision ◽  
Real Time ◽  
Cost Effective ◽  
Breast Self Examination ◽  
Cost Effective Approach ◽  
Confirmatory Tests ◽  
Evaluation Algorithm ◽  
Vision Algorithm ◽  
Self Examination

Breast cancer is the most common cancer among women worldwide and breast self-examination (BSE) is considered as the most cost-effective approach for early breast cancer detection. The general objective of this paper is to design and develop a computer vision algorithm to evaluate the BSE performance in real-time. The first stage of the algorithm presents a method for detecting and tracking the nipples in frames while a woman performs BSE; the second stage presents a method for localizing the breast region and blocks of pixels related to palpation of the breast, and the third stage focuses on detecting the palpated blocks in the breast region. The palpated blocks are highlighted at the time of BSE performance. In a correct BSE performance, all blocks must be palpated, checked, and highlighted, respectively. If any abnormality, such as masses, is detected, then this must be reported to a doctor to confirm the presence of this abnormality and proceed to perform other confirmatory tests. The experimental results have shown that the BSE evaluation algorithm presented in this paper provides robust performance.

scholarly journals Computer Vision and Machine Learning Analysis of Commercial Rice Grains: A Potential Digital Approach for Consumer Perception Studies

Sensors ◽  
2021 ◽  
Vol 21 (19) ◽  
pp. 6354
Author(s):  
Aimi Aznan ◽  
Claudia Gonzalez Viejo ◽  
Alexis Pang ◽  
Sigfredo Fuentes
Keyword(s):  
Machine Learning ◽  
Computer Vision ◽  
Imaging System ◽  
Cost Effective ◽  
Consumer Perception ◽  
Rice Grain ◽  
Ann Model ◽  
Practical Applications ◽  
Rice Samples ◽  
Artificial Neural Network Ann

Rice quality assessment is essential for meeting high-quality standards and consumer demands. However, challenges remain in developing cost-effective and rapid techniques to assess commercial rice grain quality traits. This paper presents the application of computer vision (CV) and machine learning (ML) to classify commercial rice samples based on dimensionless morphometric parameters and color parameters extracted using CV algorithms from digital images obtained from a smartphone camera. The artificial neural network (ANN) model was developed using nine morpho-colorimetric parameters to classify rice samples into 15 commercial rice types. Furthermore, the ANN models were deployed and evaluated on a different imaging system to simulate their practical applications under different conditions. Results showed that the best classification accuracy was obtained using the Bayesian Regularization (BR) algorithm of the ANN with ten hidden neurons at 91.6% (MSE = <0.01) and 88.5% (MSE = 0.01) for the training and testing stages, respectively, with an overall accuracy of 90.7% (Model 2). Deployment also showed high accuracy (93.9%) in the classification of the rice samples. The adoption by the industry of rapid, reliable, and accurate methods, such as those presented here, may allow the incorporation of different morpho-colorimetric traits in rice with consumer perception studies.

Determining Rig State from Computer Vision Analytics

2021 ◽  
Author(s):  
Crispin Chatar ◽  
Suhas Suresha ◽  
Laetitia Shao ◽  
Soumya Gupta ◽  
Indranil Roychoudhury
Keyword(s):  
Machine Learning ◽  
Computer Vision ◽  
Graphics Processing Units ◽  
New Technology ◽  
Three Dimensional ◽  
Cost Effective ◽  
Temporal Information ◽  
Sensor Data ◽  
High Tech ◽  
Drilling Rig

Abstract For years, many companies involved with drilling have searched for the ideal method to calculate the state of a drilling rig. While companies cannot agree on a standard definition of "rig state," they can agree that as we move forward in drilling optimization and with further use of remote operations and automation, that rig state calculation is mandatory in one form or the other. Internally in the service company, many methods exist for calculating rig state, but one new technology area holds promise to deliver a more efficient and cost-effective option with higher accuracy. This technology involves vision analytics. Currently, detection algorithms rely heavily on data collected by sensors installed on the rig. However, relying exclusively on sensor data is problematic because sensors are prone to failure and are expensive to maintain and install. By proposing a machine learning model that relies exclusively on videos collected on the rig floor to infer rig states, it is possible to move away from the existing methods as the industry moves to a future of high-tech rigs. Videos, in contrast to sensor data, are relatively easy to collect from small inexpensive cameras installed at strategic locations. Consequently, this paper presents machine learning pipeline that is implemented to perform rig state determination from videos captured on the rig floor of an operating rig. The pipeline can be described in two parts. Firstly, the annotation pipeline matches each frame of the video dataset to a rig state. A convolutional neural network (CNN) is used to match the time of the video with corresponding sensor data. Secondly, additional CNNs are trained, capturing both spatial and temporal information, to extract an estimation of rig state from videos. The models are trained on a dataset of 3 million frames on a cloud platform using graphics processing units (GPU). Some of the models used include a pretrained visual geometry group (VGG) network, a convolutional three-dimensional (C3D) model that used three-dimensional (3D) convolutions, and a two-stream model that uses optical flow to capture temporal information. The initial results demonstrate this pipeline to be effective in detecting rig states using computer vision analytics.

Invisible emotion magnification algorithm (IEMA) for real-time micro-expression recognition with graph-based features

2022 ◽  
Author(s):  
Adamu Muhammad Buhari ◽  
Chee-Pun Ooi ◽  
Vishnu Monn Baskaran ◽  
Raphael CW Phan ◽  
KokSheik Wong ◽  
...  
Keyword(s):  
Real Time ◽  
Expression Recognition ◽  
Micro Expression

scholarly journals A Study on Machine Learning: Elements, Characteristics and Algorithms

2018 ◽  
Vol 7 (2.19) ◽  
pp. 31
Author(s):  
K Chokkanathan ◽  
S Koteeswaran
Keyword(s):  
Machine Learning ◽  
Programming Languages ◽  
Real Time ◽  
Learning Algorithms ◽  
Cost Effective ◽  
Machine Learning Algorithms ◽  
Data Sets ◽  
Real Time Traffic ◽  
Sample Data ◽  
Spam Filters

Machine learning algorithms are used immensely for performing most important computational tasks with the help of sample data sets.  Most of the cases Machine learning algorithms will provide best solution where the programming languages failed to produce viable and economically cost-effective results.  Huge volume of deterministic problems are addressed and tackled by using the available sample data sets.  Because of this now a days machine learning concepts are extensively used in computer science and many other fields.  But still we need to explore more to implement machine learning in a specific field such as network analysis, stock trading, spam filters, traffic analysis, real-time and non-real time traffic etc., which may not be available in text books.  Here I would like to discourse some of the key points that the machine learning researchers and practitioners can make use of them.  These include shortcomings and concerns also.  

scholarly journals Quantification of toxic metallic elements using machine learning techniques and spark emission spectroscopy

2019 ◽  
Author(s):  
Seyyed Ali Davari ◽  
Anthony S. Wexler
Keyword(s):  
Machine Learning ◽  
Real Time ◽  
Emission Spectroscopy ◽  
Cost Effective ◽  
The United States ◽  
Machine Learning Techniques ◽  
Metallic Elements ◽  
Learning Techniques ◽  
Single Feature ◽  
Us Epa

Abstract. The United States Environmental Protection Agency (US EPA) list of Hazardous Air Pollutants (HAPs) includes metal elements suspected or associated with development of cancer. Traditional techniques for detecting and quantifying toxic metallic elements in the atmosphere are either not real time, hindering identification of sources, or limited by instrument costs. Spark emission spectroscopy is a promising and cost effective technique that can be used for analyzing toxic metallic elements in real time. Here, we have developed a cost-effective spark emission spectroscopy system to quantify the concentration of toxic metallic elements targeted by US EPA. Specifically, Cr, Cu, Ni, and Pb solutions were diluted and deposited on the ground electrode of the spark emission system. Least Absolute Shrinkage and Selection Operator (LASSO) was optimized and employed to detect useful features from the spark-generated plasma emissions. The optimized model was able to detect atomic emission lines along with other features to build a regression model that predicts the concentration of toxic metallic elements from the observed spectra. The limits of detections (LOD) were estimated using the detected features and compared to the traditional single-feature approach. LASSO is capable of detecting highly sensitive features in the input spectrum; however for some elements the single-feature LOD marginally outperforms LASSO LOD. The combination of low cost instruments with advanced machine learning techniques for data analysis could pave the path forward for data driven solutions to costly measurements.

scholarly journals Quantification of toxic metals using machine learning techniques and spark emission spectroscopy

2020 ◽  
Vol 13 (10) ◽  
pp. 5369-5377
Author(s):  
Seyyed Ali Davari ◽  
Anthony S. Wexler
Keyword(s):  
Machine Learning ◽  
Real Time ◽  
Toxic Metals ◽  
Emission Spectroscopy ◽  
Cost Effective ◽  
The United States ◽  
Machine Learning Techniques ◽  
Learning Techniques ◽  
Single Feature ◽  
Us Epa

Abstract. The United States Environmental Protection Agency (US EPA) list of hazardous air pollutants (HAPs) includes toxic metal suspected or associated with development of cancer. Traditional techniques for detecting and quantifying toxic metals in the atmosphere are either not real time, hindering identification of sources, or limited by instrument costs. Spark emission spectroscopy is a promising and cost-effective technique that can be used for analyzing toxic metals in real time. Here, we have developed a cost-effective spark emission spectroscopy system to quantify the concentration of toxic metals targeted by the US EPA. Specifically, Cr, Cu, Ni, and Pb solutions were diluted and deposited on the ground electrode of the spark emission system. The least absolute shrinkage and selection operator (LASSO) was optimized and employed to detect useful features from the spark-generated plasma emissions. The optimized model was able to detect atomic emission lines along with other features to build a regression model that predicts the concentration of toxic metals from the observed spectra. The limits of detections (LODs) were estimated using the detected features and compared to the traditional single-feature approach. LASSO is capable of detecting highly sensitive features in the input spectrum; however, for some toxic metals the single-feature LOD marginally outperforms LASSO LOD. The combination of low-cost instruments with advanced machine learning techniques for data analysis could pave the path forward for data-driven solutions to costly measurements.

Sign in / Sign up

Export Citation Format

Share Document