In this study, a system which describes the objects in the environment and the positions of these objects with each other has been developed. this system can be used for visually impaired people to learn what is happe...
详细信息
ISBN:
(纸本)9781538615010
In this study, a system which describes the objects in the environment and the positions of these objects with each other has been developed. this system can be used for visually impaired people to learn what is happening in their surroundings and for young children to learn objects in their surroundings. In the study, a very successful environmental information acquisition tool was developed by using the latest image segmentation, object recognition and environment description algorithms.
recognition of sign language, the main mode of communication of the hearing impaired, has attracted the attention of researchers working in the field of computervision in recent years. In this study, we propose a fas...
详细信息
Emotion recognition may be useful in any area where human and computer interacts. CNNs are known to be good at computervision tasks. However, CNNs are difficult to train, especially when the amount of data and comput...
详细信息
ISBN:
(纸本)9781538615010
Emotion recognition may be useful in any area where human and computer interacts. CNNs are known to be good at computervision tasks. However, CNNs are difficult to train, especially when the amount of data and computation power is limited. Transfer learning emerges as a cheap and efficient way of making use of pre-trained CNN classifiers. Our work has two contributions. Firstly, different CNN architectures and models trained using different datasets are investigated to find a suitable model to use in emotion recognition. Secondly, expert models for each emotion are trained. the Base model is ensembled with expert models to create a better classifier. Experiments show that our use of ensembling together with transfer learning helps to create a good classifier. Final classifier shows 68.32% accuracy on FER13 validation set.
Text detection is one of the most challenging and commonly dealt applications in computervision. Detecting text regions is the first step of the text recognition systems called Optical Character recognition. this pro...
详细信息
this paper presents a novel gesture sensing system for prosthetic limb control based on a pressure sensor array embedded in a wristband. the tendon movement which produces pressure change around the wrist can be detec...
this paper presents a novel gesture sensing system for prosthetic limb control based on a pressure sensor array embedded in a wristband. the tendon movement which produces pressure change around the wrist can be detected by pressure sensors. A microcontroller is used to gather the data from the sensors, followed by transmitting the data into a computer. A user interface is developed in LabVIEW, which presents the value of each sensor and display the waveform in real-time. Moreover, the data pattern of each gesture varies from different users due to the non-uniform subtle tendon movement. To overcome this challenge, Echo State Network (ESN), a supervised learning network, is applied to the data for calibrating different users. the results of gesture recognition show that the ESN has a good performance in multiple dimensional classifications. For experimental data collected from six participants, the proposed system classifies five gestures with an accuracy of 87.3%.
the proceedings contain 11 papers. the topics discussed include: human action recognition from RGBD videos based on retina model and local binary pattern features;a new model driven architecture for deep learning-base...
ISBN:
(纸本)9788086943428
the proceedings contain 11 papers. the topics discussed include: human action recognition from RGBD videos based on retina model and local binary pattern features;a new model driven architecture for deep learning-based multimodal lifelog retrieval;human action recognition based on 3D convolution neural networks from RGBD videos;realistic lens distortion rendering;time series social network visualization based on dimension reduction;deep learning for historical cadastral maps digitization: overview, challenges and potential;and a persistent naming system based on graph transformation rules.
In this study, the facial data of the adults and the hearing-impaired children have been analysed and compared by rule-based facial expression recognition methods. 68 face points have been selected for calculation. th...
详细信息
ISBN:
(纸本)9781538615010
In this study, the facial data of the adults and the hearing-impaired children have been analysed and compared by rule-based facial expression recognition methods. 68 face points have been selected for calculation. then using these points, the angle between the upper and down lips, the length of the corner of the lip, the angle between the eyebrows and the nose, and the angle of the aperture of the eye have been calculated. In these children, the positive facial expression can be comprehended by the lip's data and for the negative facial expression it has been determined that eyebrows, eyes and the lips should be included in the evaluation, as well. In addition to that, it has been stated that, a threshold value for these face features can not be determined within a rule based system but they can be used to determine the transitions of facial expression between different images of the same child.
Accurate and autonomous real time plant phenotyping is an essential part of modern crop monitoring and agricultural technologies. Since environmental conditions highly affect a plant's growth, accurate monitoring ...
详细信息
ISBN:
(纸本)9781538615010
Accurate and autonomous real time plant phenotyping is an essential part of modern crop monitoring and agricultural technologies. Since environmental conditions highly affect a plant's growth, accurate monitoring of phenology can a lot of information that can be used for accelerating crop production. In this paper, a deep learning architecture is utilized to recognize and classify phenological stages of several types of plants. the visual data for plants are captured every half an hour by cameras mounted on the ground agro-stations. We employ a pre-trained Convolutional Neural Network architecture (CNN) to automatically extract the features of images. the results obtained through CNN model are compared withthose obtained by employing hand crafted feature descriptors. Experimental results indicate that CNN architecture outperforms the machine learning algorithms based on hand crafted features.
Enabling computervision applications on low-power embedded systems gives rise to new challenges for embedded SW developers. Such applications implement different functionalities, like image recognition based on deep ...
详细信息
ISBN:
(纸本)9781538647561
Enabling computervision applications on low-power embedded systems gives rise to new challenges for embedded SW developers. Such applications implement different functionalities, like image recognition based on deep learning, simultaneous localization and mapping tasks. they are characterized by stringent performance constraints to guarantee real-time behaviors and, at the same time, energy constraints to save battery on the mobile platform. Even though heterogeneous embedded boards are getting pervasive for their high computational power at low power costs, they need a time consuming customization of the whole application (i.e., mapping of application blocks to CPU-GPU processing elements and their synchronization) to efficiently exploit their potentiality. Different languages and environments have been proposed for such an embedded SW customization. Nevertheless, they often find limitations on complex real cases, as their application is mutual exclusive. this paper presents a comprehensive framework that relies on a heterogeneous parallel programming model, which combines OpenMP, Pthreads, OpenVX, OpenCV, and CUDA to best exploit different levels of parallelism while guaranteeing a semi-automatic customization. the paper shows how such languages and API platforms have been interfaced, synchronized, and applied to customize an ORB-SLAM application for an NVIDIA Jetson TX2 board.
In this paper, an automatic method based on computervision for the detection of abandoned luggage/bag in public spaces is proposed. First, withthe background subtraction and foreground modeling technique, objects th...
详细信息
ISBN:
(纸本)9781538615010
In this paper, an automatic method based on computervision for the detection of abandoned luggage/bag in public spaces is proposed. First, withthe background subtraction and foreground modeling technique, objects that were not present in the previous frames and have appeared later on as a static object are detected and then whether they are luggage or not are determined using the Faster Region Proposal Convolutional Neural Network (Faster R-CNN) technique. At the same time, the person closest to the luggage is identified and defined as the owner. After that, the event analysis phase is being started. If the identified person is not in the luggage area within a certain period of time (e.g. 30 seconds), the luggage is defined as "abandoned" with a bounding box. the assessments show that the proposed approach is effective and rapid.
暂无评论