Enhanced Deep Learning Architectures for Face Liveness Detection for Static and Video Sequences

Face liveness detection is a critical preprocessing step in face recognition for avoiding face spoofing attacks, where an impostor can impersonate a valid user for authentication. While considerable research has been recently done in improving the accuracy of face liveness detection, the best current approaches use a two-step process of first applying non-linear anisotropic diffusion to the incoming image and then using a deep network for final liveness decision. Such an approach is not viable for real-time face liveness detection. We develop two end-to-end real-time solutions where nonlinear anisotropic diffusion based on an additive operator splitting scheme is first applied to an incoming static image, which enhances the edges and surface texture, and preserves the boundary locations in the real image. The diffused image is then forwarded to a pre-trained Specialized Convolutional Neural Network (SCNN) and the Inception network version 4, which identify the complex and deep features for face liveness classification. We evaluate the performance of our integrated approach using the SCNN and Inception v4 on the Replay-Attack dataset and Replay-Mobile dataset. The entire architecture is created in such a manner that, once trained, the face liveness detection can be accomplished in real-time. We achieve promising results of 96.03% and 96.21% face liveness detection accuracy with the SCNN, and 94.77% and 95.53% accuracy with the Inception v4, on the Replay-Attack, and Replay-Mobile datasets, respectively. We also develop a novel deep architecture for face liveness detection on video frames that uses the diffusion of images followed by a deep Convolutional Neural Network (CNN) and a Long Short-Term Memory (LSTM) to classify the video sequence as real or fake. Even though the use of CNN followed by LSTM is not new, combining it with diffusion (that has proven to be the best approach for single image liveness detection) is novel. Performance evaluation of our architecture on the REPLAY-ATTACK dataset gave 98.71% test accuracy and 2.77% Half Total Error Rate (HTER), and on the REPLAY-MOBILE dataset gave 95.41% accuracy and 5.28% HTER.

Download Full-text

Structural Damage Detection Based on Real-Time Vibration Signal and Convolutional Neural Network

Applied Sciences ◽

10.3390/app10144720 ◽

2020 ◽

Vol 10 (14) ◽

pp. 4720 ◽

Cited By ~ 1

Author(s):

Zhiqiang Teng ◽

Shuai Teng ◽

Jiqiao Zhang ◽

Gongfa Chen ◽

Fangsen Cui

Keyword(s):

Neural Network ◽

Experimental Data ◽

Convolutional Neural Network ◽

Real Time ◽

Structural State ◽

Detection Method ◽

Steel Frame ◽

Engineering Practice ◽

Detection Accuracy ◽

Vibration Signals

The traditional methods of structural health monitoring (SHM) have obvious disadvantages such as being time-consuming, laborious and non-synchronizing, and so on. This paper presents a novel and efficient approach to detect structural damages from real-time vibration signals via a convolutional neural network (CNN). As vibration signals (acceleration) reflect the structural response to the changes of the structural state, hence, a CNN, as a classifier, can map vibration signals to the structural state and detect structural damages. As it is difficult to obtain enough damage samples in practical engineering, finite element analysis (FEA) provides an alternative solution to this problem. In this paper, training samples for the CNN are obtained using FEA of a steel frame, and the effectiveness of the proposed detection method is evaluated by inputting the experimental data into the CNN. The results indicate that, the detection accuracy of the CNN trained using FEA data reaches 94% for damages introduced in the numerical model and 90% for damages in the real steel frame. It is demonstrated that the CNN has an ideal detection effect for both single damage and multiple damages. The combination of FEA and experimental data provides enough training and testing samples for the CNN, which improves the practicability of the CNN-based detection method in engineering practice.

Download Full-text

A novel optical image denoising technique using convolutional neural network and anisotropic diffusion for real-time surveillance applications

Journal of Real-Time Image Processing ◽

10.1007/s11554-020-01060-0 ◽

2021 ◽

Author(s):

Prabhishek Singh ◽

Achyut Shankar

Keyword(s):

Neural Network ◽

Convolutional Neural Network ◽

Real Time ◽

Image Denoising ◽

Anisotropic Diffusion ◽

Optical Image ◽

Surveillance Applications

Download Full-text

Intent Detection-Based Lithuanian Chatbot Created via Automatic DNN Hyper-Parameter Optimization

Frontiers in Artificial Intelligence and Applications - Human Language Technologies – The Baltic Perspective ◽

10.3233/faia200608 ◽

2020 ◽

Cited By ~ 1

Author(s):

Jurgita Kapočiūtė-Dzikienė

Keyword(s):

Neural Network ◽

Error Analysis ◽

Natural Language ◽

Convolutional Neural Network ◽

Parameter Optimization ◽

Deep Neural Network ◽

Short Term Memory ◽

Detection Accuracy ◽

Term Memory ◽

Long Short Term Memory

In this paper, we tackle an intent detection problem for the Lithuanian language with the real supervised data. Our main focus is on the enhancement of the Natural Language Understanding (NLU) module, responsible for the comprehension of user’s questions. The NLU model is trained with a properly selected word vectorization type and Deep Neural Network (DNN) classifier. During our experiments, we have experimentally investigated fastText and BERT embeddings. Besides, we have automatically optimized different architectures and hyper-parameters of the following DNN approaches: Long Short-Term Memory (LSTM), Bidirectional LSTM (BiLSTM) and Convolutional Neural Network (CNN). The highest accuracy=∼0.715 (∼0.675 and ∼0.625 over random and majority baselines, respectively) was achieved with the CNN classifier applied on a top of BERT embeddings. The detailed error analysis revealed that prediction accuracies degrade for the least covered intents and due to intent ambiguities; therefore, in the future, we are planning to make necessary adjustments to boost the intent detection accuracy for the Lithuanian language even more.

Download Full-text

Multi-collaborative Regression Convolutional Neural Network for Vehicle Angle Detection

Recent Advances in Electrical & Electronic Engineering (Formerly Recent Patents on Electrical & Electronic Engineering) ◽

10.2174/2352096513999200628100357 ◽

2020 ◽

Vol 13 (8) ◽

pp. 1227-1241

Author(s):

Guoqiang Chen ◽

Mengchao Liu ◽

Hongpeng Zhou ◽

Bingxin Bai

Keyword(s):

Neural Network ◽

Convolutional Neural Network ◽

Real Time ◽

Autonomous Vehicles ◽

Training Sample ◽

Detection Accuracy ◽

Detection Error ◽

Detection Failure ◽

Pose Detection ◽

Parking Lot

Background: The vehicle pose detection plays an important role in monitoring vehicle behavior and the parking situation. The real-time detection of vehicle pose with high accuracy is of great importance. Objective: The goal of the work is to construct a new network to detect the vehicle angle based on the regression Convolutional Neural Network (CNN). The main contribution is that several traditional regression CNNs are combined as the Multi-Collaborative Regression CNN (MCR-CNN), which greatly enhances the vehicle angle detection precision and eliminates the abnormal detection error. Methods: Two challenges with respect to the traditional regression CNN have been revealed in detecting the vehicle pose angle. The first challenge is the detection failure resulting from the conversion of the periodic angle to the linear angle, while the second is the big detection error if the training sample value is very small. An MCR-CNN is proposed to solve the first challenge. And a 2- stage method is proposed to solve the second challenge. The architecture of the MCR-CNN is designed in detail. After the training and testing data sets are constructed, the MCR-CNN is trained and tested for vehicle angle detection. Results: The experimental results show that the testing samples with the error below 4° account for 95% of the total testing samples based on the proposed MCR-CNN. The MCR-CNN has significant advantages over the traditional vehicle pose detection method. Conclusion: The proposed MCR-CNN cannot only detect the vehicle angle in real-time, but also has a very high detection accuracy and robustness. The proposed approach can be used for autonomous vehicles and monitoring of the parking lot.

Download Full-text

High-Speed Ship Detection in SAR Images Based on a Grid Convolutional Neural Network

Remote Sensing ◽

10.3390/rs11101206 ◽

2019 ◽

Vol 11 (10) ◽

pp. 1206 ◽

Cited By ~ 16

Author(s):

Tianwen Zhang ◽

Xiaoling Zhang

Keyword(s):

Neural Network ◽

Convolutional Neural Network ◽

Real Time ◽

High Speed ◽

Earth Observation ◽

Detection Accuracy ◽

Processing Unit ◽

Sar Images ◽

Ship Detection ◽

Detection Speed

As an active microwave sensor, synthetic aperture radar (SAR) has the characteristic of all-day and all-weather earth observation, which has become one of the most important means for high-resolution earth observation and global resource management. Ship detection in SAR images is also playing an increasingly important role in ocean observation and disaster relief. Nowadays, both traditional feature extraction methods and deep learning (DL) methods almost focus on improving ship detection accuracy, and the detection speed is neglected. However, the speed of SAR ship detection is extraordinarily significant, especially in real-time maritime rescue and emergency military decision-making. In order to solve this problem, this paper proposes a novel approach for high-speed ship detection in SAR images based on a grid convolutional neural network (G-CNN). This method improves the detection speed by meshing the input image, inspired by the basic thought of you only look once (YOLO), and using depthwise separable convolution. G-CNN is a brand new network structure proposed by us and it is mainly composed of a backbone convolutional neural network (B-CNN) and a detection convolutional neural network (D-CNN). First, SAR images to be detected are divided into grid cells and each grid cell is responsible for detection of specific ships. Then, the whole image is input into B-CNN to extract features. Finally, ship detection is completed in D-CNN under three scales. We experimented on an open SAR Ship Detection Dataset (SSDD) used by many other scholars and then validated the migration ability of G-CNN on two SAR images from RadarSat-1 and Gaofen-3. The experimental results show that the detection speed of our proposed method is faster than the existing other methods, such as faster-regions convolutional neural network (Faster R-CNN), single shot multi-box detector (SSD), and YOLO, under the same hardware environment with NVIDIA GTX1080 graphics processing unit (GPU) and the detection accuracy is kept within an acceptable range. Our proposed G-CNN ship detection system has great application values in real-time maritime disaster rescue and emergency military strategy formulation.

Download Full-text

Real-time fault diagnosis using deep fusion of features extracted by parallel long short-term memory with peephole and convolutional neural network

Proceedings of the Institution of Mechanical Engineers Part I Journal of Systems and Control Engineering ◽

10.1177/0959651820948291 ◽

2020 ◽

pp. 095965182094829

Author(s):

Funa Zhou ◽

Zhiqiang Zhang ◽

Danmin Chen

Keyword(s):

Neural Network ◽

Fault Diagnosis ◽

Convolutional Neural Network ◽

Real Time ◽

Short Term Memory ◽

Two Dimensional ◽

Short Term ◽

Term Memory ◽

One Dimensional ◽

Long Short Term Memory

Analysis of one-dimensional vibration signals is the most common method used for safety analysis and health monitoring of rotary machines. How to effectively extract features involved in one-dimensional sequence data is crucial for the accuracy of real-time fault diagnosis. This article aims to develop more effective means of extracting useful features potentially involved in one-dimensional vibration signals. First, an improved parallel long short-term memory called parallel long short-term memory with peephole is designed by adding a peephole connection before each forget gate to prevent useless information transferring in the cell. It can not only solve the memory bottleneck problem of traditional long short-term memory for long sequence but also can make full use of all possible information helpful for feature extraction. Second, a fusion network with new training mechanism is designed to fuse features extracted from parallel long short-term memory with peephole and convolutional neural network, respectively. The fusion network can incorporate two-dimensional screenshot image into comprehensive feature extraction. It can provide more accurate fault diagnosis result since two-dimensional screenshot image is another form of expression for one-dimensional vibration sequence involving additional trend and locality information. Finally, real-time two-dimensional screenshot image is fed into convolutional neural network to secure a real-time online diagnosis which is the primary requirement of the engineers in health monitoring. Validity of the proposed method is verified by fault diagnosis for rolling bearing and gearbox.

Download Full-text

Real-time Respiratory Tumor Motion Prediction based on Temporal Convolutional Neural Network (Preprint)

10.2196/preprints.27235 ◽

2021 ◽

Author(s):

Panchun Chang ◽

Jun Dang ◽

Jianrong Dai ◽

Wenzheng Sun

Keyword(s):

Neural Network ◽

Convolutional Neural Network ◽

Real Time ◽

Target Location ◽

Short Term Memory ◽

Radiation Beam ◽

Dynamic Tracking ◽

Time Latency ◽

Beam Delivery ◽

Deep Learning Model

BACKGROUND Dynamic tracking of tumor with radiation beam in radiation therapy requires prediction of real-time target location ahead of beam delivery as the treatment with beam or gating tracking brings in time latency. OBJECTIVE A deep learning model based on a temporal convolutional neural network (TCN) using multiple external makers was developed to predict internal target location through multiple external markers in this study. METHODS The respiratory signals from 69 treatment fractions of 21 cancer patients treated with the Cyberknife Synchrony device were used to train and test the model. The reported model’s performance was evaluated through comparing with a long short term memory model in terms of root-mean-square-error (RMSE) between real and predicted respiratory signals. Besides, the effect of external marker number was also investigated. RESULTS The average RMSEs (mm) for 480-ms ahead of prediction using TCN model in the superior–inferior (SI), anterior–posterior (AP) and left–right (LR) and radial directions were 0.49, 0.28, 0.25 and 0.67, respectively. CONCLUSIONS The experiment results demonstrated that the TCN respiratory prediction model could predict the respiratory signals with sub-millimeter accuracy.

Download Full-text