Recently, a number of methods for dynamic hand gesture recognition has been proposed.However, deployment such in practical application still to face with many challenges due the variation view point, complex background or subject style.In this work, we deeply investigate performance advanced convolutional neural networks specific case gestures and evaluate how robust it is above variations.To end, adopt an existing 3D network which was originally proposed general human action obtained very...

This paper presents a new method for hand segmentation from images and video. The based mainly on an advanced technique instance (Mask RCNN) which has been shown very efficient in task COCO dataset. However, Mask R-CNN some limitations. It works still images, so cannot explore temporal information of the object interest such as dynamic gestures. Second usually fails to detect suffered motion blur at low resolution hand. Our proposed improves by integrating Mean Shift tracker that tracks...

Dynamic hand gesture recognition is a challenge field evenly this topic has been studied for long time because of lack feasible techniques deployed Human-Computer Interaction (HCI) applications. In paper, we propose new type gestures which presents cyclic pattern shapes during movement. Through mapping commands (e.g., turn devices on/off; increasing volume/channel) as output system, main purposes the proposed are to provide natural and way in control alliances smart home such television,...

Land consolidation is an effective solution for the hindrances in agricultural production and rural development caused by land fragmentation. In Red River Delta of Vietnam, where still highly fragmented, application required. By using a bottom-up approach, paper aims to clarify effect on farm households selected communities (as case studies) two provinces (Hung Yen Vinh Phuc) Delta. With primary structured semi-structured interview method, 172 household questionnaires 22 in-depth (from local...

Deep neural networks (DNNs) have made outstanding achievements in a wide variety of domains. For deep learning tasks, large enough datasets are required for training efficient DNN models. However, big not always available, and they costly to build. Therefore, balanced solutions model efficiency data size caught the attention researchers recently. Transfer techniques most common this. In transfer learning, is pre-trained on dataset then applied new task with modest data. This fine-tuning...

Phase synchronization issue, that is caused by spotting gestures from video stream, varying frame-rates, speed of subject's implementation, should be overcome in developing Human-Computer Interaction (HCI) application using dynamic hand gestures. This paper tackles an interpolation technique to efficiently solve this issue. We firstly propose a new representation space consists both spatial and temporal features extracted the The are based on manifold learning (ISOMAP) takes into account...

In this paper, we tackle advantages of cyclical movement patterns hand gestures. The are defined as closed-form which moves away from a rest position, follows one or more series the shapes and returns to its position. Due pattern characteristic, phase gestures supportive cues for deploying robust recognition schemes. We conduct spatial-temporal representation takes into account both movements during gesture. alignment then is deployed in conducted space. proposed scheme ensures inter-period...

Hand gesture recognition has attracted the attention of many scientists, because its high applicability in fields such as sign language expression and human machine interaction. Many approaches have been deployed to detect recognize hand gestures, like wearable devices, image information, and/or a combination sensors computer vision. However, method using brings much higher accuracy is less affected by occlusion, lighting conditions, complex background. Existing solutions separately utilize...

<p>In this study, we extensively analyze and evaluate the performance of recent deep neural networks (DNNs) for hand gesture recognition static gestures in particular. To end, captured an unconstrained dataset with complex appearances, shapes, scales, backgrounds, viewpoints. We then deployed some new trending convolution neuron (CNNs) classification. arrived at three major conclusions: i) DenseNet121 architecture is best rate through almost evaluated red, green, blue (RGB)...

Person re-identification (ReID), a critical task in surveillance systems, has obtained impressive advances recent years. However, most current works focus on improving the person accuracy. In practical terms, direct use of these seems difficult, even infeasible. Among features proposed for representation ReID, Gaussian (GOG) been proved to be robust. Towards applying this feature usage, work, we simultaneously propose two improvements. First, re-implement and perform intensive experiments...

Human action recognition (HAR) under different camera viewpoints is the most critical requirement for practical deployment. In this paper, we propose a novel method that leverages successful deep learning-based features representation and multi-view analysis to accomplish robust HAR viewpoint changes. Specifically, investigate various learning techniques, from 2D CNNs 3D capture spatial temporal characteristics of actions at each separated view. A common feature space then constructed keep...

This paper argues that an user-guide plays important role to make a robust and real-time hand posture recognition system. Instead of designing new algorithm, we propose scheme which handles issues environmental conditions as well appearance-based features for detections. guide estimates heuristic parameters whose values strongly affect the results. The experimental results confirm even by utilizing simple proposed method significantly improves rate. Without training end-user, rate achieves...

Human gesture recognition is an attractive research area in computer vision with many applications such as human-machine interaction, virtual reality, etc. Recent deep learning techniques have been efficiently applied for recognition, but they require a large and diverse amount of training data. In fact, the available datasets contain mostly static gestures and/or certain fixed viewpoints. Some dynamic gestures, are not poses this paper, we propose novel end-to-end framework from unknown It...

<p>Deep learning models have been successfully applied to many visual tasks. However, they tend be increasingly cumbersome due their high computational complexity and large storage requirements. How compress convolutional neural network (CNN) while still maintain efficiency has received increasing attention from the community, knowledge distillation (KD) is efficient way do this. Existing KD methods focused on selection of good teachers multiple teachers, or layers, which cumbersome,...

Deep Neural Networks (DNNs) have become a promising solution for detecting abnormal human behaviors. However, building an efficient DNN model in terms of both computational cost and classification accuracy is still challenging problem. Furthermore, there are limited existing datasets behavior detection, each focuses on certain context. Therefore, trained dataset will be adaptive particular context not suitable others. This study proposes framework with attention Knowledge Distillation (KD)...

Multi-modal or multi-view dataset that was captured from various resources (e.g. RGB and Depth) of a subject at the same time. Combination between different cues has still faced to many challenges as unique data complementary in-formation. In adition, proposed method for multiple modalities recognition consists discrete blocks, such as: extract features separative flows, combine features, classify gestures. To address challenges, we pro-posed two novel end-to-end hand posture frameworks,...

Video or image-based people counting in real-time has multiple applications intelligent transportation, density estimation class management, and so on. This problem is usually carried out by detecting using conventional detectors. However, this approach can be failed when stay various postures are occluded each other. In paper, we notice that even a main part of human body occluded, their face head still observable. We then propose method counts based on detection pairing. Instead deploying...

Recently, license plate recognition has been become an attractive field in computer vision. Which consists some main steps such as: data collection, detection, character separation, segmentation, characters and series connection. Many state-of-the-art methods have proposed while almost these approaches utilize complex algorithm. That spends a large time cost to obtain competitive accuracy; and/or high equipment performance as CPU, GPU, cameras so on. In addition, recent not deployed...