A Fusion Approach Based on HOG and Adaboost Algorithm for Face Detection under Low- Resolution Images

Modern face recognition systems extract face representations using deep neural networks (DNNs) and give excellent identification and verification results, when tested on high resolution (HR) images. However, the performance of such an algorithm degrades significantly for low resolution (LR) images. A straight forward solution could be to train a DNN, using simultaneously, high and low resolution face images. This approach yields a definite improvement at lower resolutions but suffers a performance degradation for high resolution images. To overcome this shortcoming, we propose to train a network using both HR and LR images under the guidance of a fixed network, pretrained on HR face images. The guidance is provided by minimising the KL-divergence between the output Softmax probabilities of the pretrained (i.e., Teacher) and trainable (i.e., Student) network as well as by sharing the Softmax weights between the two networks. The resulting solution is tested on down-sampled images from FaceScrub and MegaFace datasets and shows a consistent performance improvement across various resolutions. We also tested our proposed solution on standard LR benchmarks such as TinyFace and SCFace. Our algorithm consistently outperforms the state-of-the-art methods on these datasets, confirming the effectiveness and merits of the proposed method.

Fatigue driving detection helps reduce the occurrence of traffic accidents. Currently, the fatigue detection methods still have following challenges: 1) In the visual fatigue detection system, face detection is the foundation of fatigue detection and it is always susceptible to changes of illumination and facial expression, which lowers the accuracy of face detection. 2) It is so difficult to distinguish some behaviors of a driver such as singing and yawning in a single frame that making false fatigue detection in the visual fatigue detection system. To overcome the problems mentioned above, a fatigue detection method based on sequential mouth feature with long short-term memory network is proposed. The proposed method mainly consists of two parts: locating of facial landmark based on face and mouth detection, and yawning detection based on residual network (Resnet) and long short-term memory (LSTM). The face detection method is based on Histogram of Oriented Gradient (HOG) feature and Support Vector Machine (SVM) classifier, which is robust to the changes of illumination and slightly facial expression. The mouth detection is based on cascaded regression tree, which extracts the mouth region from the detected face. The Resnet and LSTM based yawning detection extracts spatial-temporal features from the mouth regions between frames and distinguishes the behaviors about yawning to identify fatigue of a driver with sequential mouth feature. The proposed method achieves accuracy of 94.9% on the YawDD dataset, which outperforms the compared methods. Experimental results demonstrate that the proposed method satisfies the performance of fatigue detection.

The histogram of oriented gradient(HOG) is a feature descriptor used for object detection in the computer vision and image processing, and it is widely used for pedestrian detection. The conspicuous image feature can improve the detective accuracy of the pedestrian detection. In order to improve the conspicuousness of extracted gradient, this paper modifies the gradient extraction operator based on the traditional HOG algorithm. By the tests of different operators, this paper chooses the Prewitt operator to extract the gradient information. Experimental results indicate that the mean and variance of extracted gradient are larger than the gradient of traditional HOG algorithm. The extracted gradient should be generated the conspicuous HOG feature that may improve the performance of pedestrian detection.

The estimation of the number of people in surveillance areas is essential for monitoring crowded scenes. When density of a zone increases to a certain approximated level, people's safety can be endangered. Detection of human is a prerequisite for density estimation, tracking, activity recognition and anomaly detection even in non congested areas. This paper presents a robust hybrid approach for face detection in crowd by combining the skin color segmentation and a Histogram of Oriented Gradients(HOG) with Support Vector Machine(SVM) architecture. Initially, image enhancement is performed to improve the detection rate. An edge preserving pyramidal approach is applied for multiscale representation of an image. Skin color segmentation is done with combination of YCbCr and RGB color model, and HOG features are extracted from the segmented skin region. We trained the SVM classifier by Muct and FEI databases which consist 751 and 2800 face images respectively. The accuracy of this approach is evaluated by testing it on BAO multiple face database and on various manually collected images captured in surveillance areas. Experimental results demonstrate that the supplementary skin color segmentation with HOG is more potent for increasing the detection rate than using HOG features only. The proposed approach achieves 98.02% accuracy which is higher in comparison to Viola Jones and fast face detection method.

Face detection by low-resolution image (LR) is one of the key aspects of Human-Computer Interaction(HCI). Due to the LR image, which has changes in pose, lighting, and illumination, the performance of face recognition is reduced. In this work, we propose the Deep Belief Network-Crossover based Firefly (DBN-CROFF) method for face recognition from low-resolution images. The Histogram of Gradient (HOG) and 2-Dimensional Discrete Wavelet Transform (2D-DWT) to extract facial width, size of the cheeks, skin tone, nose, and lip shape features from facial data. The Kernel Principle Component Analysis (k-PCA) is used to successfully reduce the dimension of the feature. The experimental performance of the proposed method is evaluated using four datasets namely LFW, Multi-PIE, Extended Yale-B, and FERET with conventional techniques. Finally, the proposed DBN-CROFF solution surpasses the other conventional facial recognition approaches by giving a higher accuracy of recognization.

Principal curvatures based rotation invariant algorithms for efficient texture classification

This paper presents a real-time face detection system using a moving camera. The proposed system consists of three modules, including (1) detection of face candidates: Face candidates are generated using the information of skin color, edges, and face area, (2) verification of face candidates: HOG (Histogram of Oriented Gradient) features are generated from face candidates and a two-class C-SVM (Support Vector Machine) classifier with pretrained face samples is employed to determine whether face candidates are real faces or not, (3) face tracking: Overlapping area of two face targets in current and previous frames is estimated to determine whether the tracking will be continuous or not. By use of estimation of face size, the proposed method can avoid a huge amount of computation time that is required by a point-by-point scanning way in conventional methods. Moreover, the accuracy of the face detection can be improved greatly. The proposed system can successfully detect most faces of the crowds in open space, which is beneficial for quickly searching the specified persons to prevent the occurrence of possible criminal events.

Forest fires solely are responsible for a multitude of environmental disasters. This in turn results in economic and ecological/environmental damage, thus bringing peril to people’s lives. Automatic surveillance vehicles and early detection in forest fires have gained more importance in comparison to the traditional human monitoring. This is because human monitoring is subjective, which affects reliability in detection and is one of the major concerns in forest-fire detection systems. In the present system, the process is time-consuming and false alarms must be declared manually. This research is an attempt to solve the aforementioned problem. The hypothesis is to deploy a drone equipped with a thermal camera to capture the area on fire and locate victims and give the shortest distance to them. This process is entirely done when the firefighters are in transit to the location. The main focus of this text is thermal video processing since it plays a crucial role in the entire prototype. This paper highlights the difference between two human detection algorithms: Histogram of Oriented Gradients (HOG) [1] and You Only Look Once (YOLOv2) [2, 3] for thermal videos. Furthermore, an algorithm for detecting fire in a thermal video has also been implemented. The inputs are various thermal videos which are available publicly. The conclusion is that the HOG algorithm is faster than the YOLO algorithm for thermal inputs but the latter is more accurate.KeywordsThermal video processingYOLOv2HOGFire detectionHuman detectionSurveillance

Face recognition has attracted enormous interest from researchers due to its widescale adaptation. However, most of the work in the literature is based on face recognition in controlled environments i.e., high-resolution face images with slight variation in pose, illumination and expression. CNN is used as backbone architecture in the face recognition models. Low-resolution face images having large variations in pose, illumination and expressions make a face recognition problem more challenging. It is difficult to extract discriminative features from low-resolution images. That's why the performance of face recognition models trained on high-resolution images deteriorate on low-resolution images. In this work, quality of low-resolution images is enhanced through multiple super-resolution GANs in order to improve the identification accuracy. Experiments are performed on SCface dataset and the improvement in the identification accuracy shows the effectiveness of using enhanced images generated through super-resolution GANs.

High false positive face detection is a crucial problem which leads to low performance face recognition in surveillance system. The performance can be increased by reducing these false positives so that non-face can be discarded first prior to recognition. This paper presents a combination of two well known algorithms, Adaboost and Neural Network, to detect face in static images which is able to reduce the false-positives drastically. This method utilizes Haar-like features to extract the face rapidly using integral image. A cascade Adaboost classifier is used to increase the face detection speed. Due to using only this cascade Adaboost produces high false-positives, neural network is used as the final classifier to verify face or non-face. For a faster processing time, hierarchical Neural Network is used to increase the face detection rate. Experiments on four different face databases, which consist more than one thousand images, have been conducted. Results reveal that the proposed method achieves about 93.34% of detection rate and 0.34% of false-positives compared to original cascade Adaboost method which achieves about 98.13% of detection rate with 6.50% of false-positives. The processed images size is 240 × 320 pixels. Each frame is processed at about 2.25 sec which is slightslower than the original method, which only takes about 0.82 sec.

This paper proposed a novel method for face hallucination based on a neighbor embedding technique. Traditional neighbor embedding approaches often offer counterintuitive results because consistency between high resolution images and low resolution images cannot be preserved without taking the intrinsic features of the image patches into account. In order to reinforce the consistency, on the one hand, we exploit the singular-values-unique (SVU) features inspired by singular values decomposition (SVD) successfully applied in image processing. On the other hand, we introduced the Histograms of Oriented Gradients (HOG) features to characterize the local geometric structure of the image patches to alleviate the effects of noise. At last, the learning space is extended to a coupled feature space that combines the SVU and HOG features. Simulation experiments show that this proposed approach could provide competitive results in simulation experiments in subjective and objective quality.

Object detection in images is a computing demanding task which usually needs to deal with the detection of different classes of objects, and thus requiring variations and adaptations easily provided by software solutions. Object detection algorithms are being part of real-time smarter embedded systems, such as automotive, medical, robotics and security systems. In most embedded systems, efficient implementations of object oriented algorithms need to provide high performance, low power consumption, and programmability to allow greater development flexibility. The Histogram of Oriented Gradients (HOG) is one of the most widely used algorithms for object detection in images. In this paper, we show our work towards mapping the HOG algorithm to an FPGA-based system consisting of multiple Nios II softcore processors and bearing in mind high-performance and programmability issues. We show how to reduce 19x the algorithms execution time by source to source transformations and specially avoiding redundant processing. Furthermore, we show how the use of pipelining processing using three Nios II processors allows a speedup of 49x compared to the embedded baseline application.

In this paper, a new method for facial expression classification is proposed, which uses the histograms of oriented gradients (HOG) algorithm to extract facial expression features and the sparse representation classifier (SRC) to classify facial expressions with a large variation of poses. The HOG algorithm was selected due to its effectiveness in picking up both local and global facial expression features in different orientations and scales, and the SRC was chosen because of its proven effectiveness in face recognition. A novelty of the proposed approach is that given a facial image for classification, its pose is determined first to select a pose-dependent dictionary for the SRC procedure. The paper also discusses ways of selecting parameters for improving the effectiveness of the HOG algorithm. The proposed method was applied to two multi-pose facial expression databases and satisfactory results were obtained for the majority of facial expressions under various poses.

Histogram of Oriented Gradients (HOG) feature descriptor is very effective to represent objects and is widely used in human and face detection. In this paper, HOG feature descriptor is applied for Content-based Image Retrieval (CBIR). For handling similarity measurement of large amount of database, vocabulary tree is used. Experimental results illustrate the comparative analysis of retrieval system based on HOG feature descriptor and Gabor transform feature descriptor. It is verified that HOG-based retrieval system improves Average Precision (AP) and Average Recall (AR) (56.75% and 38.45%, respectively) from Gabor-transform-based retrieval system (41.20% and 25.41%, respectively). All the experiments are performed on Corel 1000 natural image database.

In order to detect pedestrians in many applications, histogram of oriented gradient (HOG) feature is widely used. Although HOG feature can provide high detection accuracy, fast detection time is hardly achieved due to its computational complexity. In this paper, a new method for pedestrian detection using HOG feature is proposed to achieve both high detection rate and fast detection time. Multi-scale cells and blocks of HOG feature are used to improve detection accuracy, and they are computed by using integral HOG to improve detection speed. In order to further improve detection quality in terms of accuracy and speed, two-stage approach is adopted in the proposed method. Experimental results show that the proposed method improves both detection rate and speed.