Past studies used human rated as the way of establishing a corpus which costs a lot of time and money but contains insufficient words, also the Categorical Approach was seldom used for building corpus, which may also lead to study bias. Therefore, study 1 of present study has used the Spreading Activation Model as the structure, and used big data of text corpus and word co-occurrences to build a corpus that contains more categories of emotions and much more words. First, study 1 selected the words that can clearly describe the meanings or can effectively evoke the feeling of its emotion category for seven emotions, including Happiness, Surprise, Sadness, Anger, Disgust, Fear, and Love. Then study 1 calculated the averages of co-occurrences for selected words and text corpora by seven emotions categories (measure is Baroni-Urbani, unit is chunk), it computes the averages of co-occurrences by emotional categories for 33669 words, it represents the conceptual consonance of words and the emotions. Study 2 has investigated the practical use of the corpus built in study 1, and used C-LIWC dictionary which was built by human rated as a comparison, taking the posts of Happy Board, Sad Board, Hate Board of PTT Bulletin Board System into the analyses of emotions recognition, result showed that Corpus of 7 Types Emotion Co-occurrences Words" built in study 1 had higher correct rate than human rated corpus. Present study has also compared the correct rates between the Corpus of 7 Types Emotion Co-occurrences Words and CLIWC (Chinese Linguistic Inquiry and Word Count), result showed correct rates of two databases were significant different, the corpus of present study has higher correct rate. Present study has built a text corpus for the material of emotion research, and the results also supports a potential of building the corpora of emotional words with big data measures.

Recently, a series of works in computer vision have shown promising results on various image and video understanding tasks using self-attention. However, due to the quadratic computational and memory complexities of self-attention, these works either apply attention only to low-resolution feature maps in later stages of a deep network or restrict the receptive field of attention in each layer to a small local region. To overcome these limitations, this work introduces a new global self-attention module, referred to as the GSA module, which is efficient enough to serve as the backbone component of a deep network. This module consists of two parallel layers: a content attention layer that attends to pixels based only on their content and a positional attention layer that attends to pixels based on their spatial locations. The output of this module is the sum of the outputs of the two layers. Based on the proposed GSA module, we introduce new standalone global attention-based deep networks that use GSA modules instead of convolutions to model pixel interactions. Due to the global extent of the proposed GSA module, a GSA network has the ability to model long-range pixel interactions throughout the network. Our experimental results show that GSA networks outperform the corresponding convolution-based networks significantly on the CIFAR-100 and ImageNet datasets while using less parameters and computations. The proposed GSA networks also outperform various existing attention-based networks on the ImageNet dataset.

Mutual gaze detection, i.e., predicting whether or not two people are looking at each other, plays an important role in understanding human interactions. In this work, we focus on the task of image-based mutual gaze detection, and propose a simple and effective approach to boost the performance by using an auxiliary 3D gaze estimation task during the training phase. We achieve the performance boost without additional labeling cost by training the 3D gaze estimation branch using pseudo 3D gaze labels deduced from mutual gaze labels. By sharing the head image encoder between the 3D gaze estimation and the mutual gaze detection branches, we achieve better head features than learned by training the mutual gaze detection branch alone. Experimental results on three image datasets show that the proposed approach improves the detection performance significantly without additional annotations. This work also introduces a new image dataset that consists of 33.1K pairs of humans annotated with mutual gaze labels in 29.2K images.

With the advent of the 5G and Artificial Intelligence of Things (AIoT) era, related technologies such as the Internet of Things, big data analysis, cloud applications, and artificial intelligence have brought broad prospects to many application fields, such as smart homes, autonomous vehicles, smart cities, healthcare, and smart campus. At present, most university campus app is presented in the form of static web pages or app menus. This study mainly developed a Deep Neural Network (DNN) based emotionally aware campus virtual assistant. The main contributions of this research are: (1) This study introduces the Chinese Word Embedding to the robot dialogue system, effectively improving dialogue tolerance and semantic interpretation. (2) The traditional method of emotion identification must first tokenize the Chinese sentence, analyze the clauses and part of speech, and capture the emotional keywords before being interpreted by the expert system. Different from the traditional method, this study classifies the input directly through the convolutional neural network after the input sentence is converted into a spectrogram by Fourier Transform. (3) This study is presented in App mode, which is easier to use and economical. (4) This system provides a simple voice response interface, without the need for users to find information in complex web pages or app menus.

Highly expressed in cancer 1 (Hec1) plays an essential role in mitosis and is correlated with cancer formation, progression, and survival. Phosphorylation of Hec1 by Nek2 kinase is essential for its mitotic function, thus any disruption of Hec1/Nek2 protein-protein interaction has potential for cancer therapy. We have developed T-1101 tosylate (9j tosylate, 9j formerly known as TAI-95), optimized from 4-aryl-N-pyridinylcarbonyl-2-aminothiazole of scaffold 9 by introducing various C-4' substituents to enhance potency and water solubility, as a first-in-class oral clinical candidate for Hec1 inhibition with potential for cancer therapy. T-1101 has good oral absorption, along with potent in vitro antiproliferative activity (IC50: 14.8-21.5 nM). It can achieve high concentrations in Huh-7 and MDA-MB-231 tumor tissues, and showed promise in antitumor activity in mice bearing human tumor xenografts of liver cancer (Huh-7), as well as of breast cancer (BT474, MDA-MB-231, and MCF7) with oral administration. Oral co-administration of T-1101 halved the dose of sorafenib (25 mg/kg to 12.5 mg/kg) required to exhibit comparable in vivo activity towards Huh-7 xenografts. Cellular events resulting from Hec1/Nek2 inhibition with T-1101 treatment include Nek2 degradation, chromosomal misalignment, and apoptotic cell death. A combination of T-1101 with either of doxorubicin, paclitaxel, and topotecan in select cancer cells also resulted in synergistic effects. Inactivity of T-1101 on non-cancerous cells, a panel of kinases, and hERG demonstrates cancer specificity, target specificity, and cardiac safety, respectively. Subsequent salt screening showed that T-1101 tosylate has good oral AUC (62.5 mu M.h), bioavailability (F = 77.4%), and thermal stability. T-1101 tosylate is currently in phase I clinical trials as an orally administered drug for cancer therapy. (c) 2020 Elsevier Masson SAS. All rights reserved.

Although deep learning approaches have achieved performance surpassing humans for still image-based face recognition, unconstrained video-based face recognition is still a challenging task due to large volume of data to be processed and intra/inter-video variations on pose, illumination, occlusion, scene, blur, video quality, etc. In this work, we consider challenging scenarios for unconstrained video-based face recognition from multiple-shot videos and surveillance videos with low-quality frames. To handle these problems, we propose a robust and efficient system for unconstrained video-based face recognition, which is composed of modules for face/fiducial detection, face association, and face recognition. First, we use multi-scale single-shot face detectors to efficiently localize faces in videos. The detected faces are then grouped respectively through carefully designed face association methods, especially for multi-shot videos. Finally, the faces are recognized by the proposed face matcher based on an unsupervised subspace learning approach and a subspace-to-subspace similarity metric. Extensive experiments on challenging video datasets, such as Multiple Biometric Grand Challenge (MBGC), Face and Ocular Challenge Series (FOCS), IARPA Janus Surveillance Video Benchmark (IJB-S) for low-quality surveillance videos and IARPA JANUS Benchmark B (IJB-B) for multiple-shot videos, demonstrate that the proposed system can accurately detect and associate faces from unconstrained videos and effectively learn robust and discriminative features for recognition.

In this paper, we present an end-to-end system for the unconstrained face verification problem based on deep convolutional neural networks (DCNN). The end-to-end system consists of three modules for face detection, alignment and verification and is evaluated using the newly released IARPA Janus Benchmark A (IJB-A) dataset and its extended version Janus Challenging set 2 (JANUS CS2) dataset. The IJB-A and CS2 datasets include real-world unconstrained faces of 500 subjects with significant pose and illumination variations which are much harder than the Labeled Faces in the Wild (LFW) and Youtube Face (YTF) datasets. Results of experimental evaluations for the proposed system on the IJB-A dataset are provided.

In this paper, we present a new video-based face identification algorithm, where the target (i.e., person of interest) in the probe video is only annotated once with a face bounding box in a frame and the video may consist of multiple shots. Most video face identification techniques assume that the video is of single shot, and thus the bounding boxes of the target face can be extracted by tracking a face across the video frames. Nevertheless, such automatic annotation is vulnerable to the drifting of the face tracker, and the face tracking algorithm is inadequate to associate the face images of the target across multiple shots. In this paper, we propose a target face association (TFA) technique that retrieves a set of representative face images in a given video that are likely to have the same identity as the target face. These face images are then utilized to construct a robust face representation of the target face for searching the corresponding subject in the gallery. Since two faces that appear in the same video frame cannot belong to the same person, such cannot-link constraints are utilized for learning a target-specific linear classifier for establishing the intra/inter-shot face association of the target. Experimental results on the newly released JANUS challenge set 3 (JANUS CS3) dataset show that our method generates robust representations from target-annotated videos and demonstrates good performance for the task of video-based face identification problem.

Face recognition in outdoor environment is a challenging task due to illumination changes, pose variations, and occlusions. In this chapter, we discuss several techniques for continuous tracking of faces acquired by an outdoor camera network as well as a face matching algorithm. Active camera networks are capable of reconfiguring the camera parameters to collaboratively capture the close-up views of face images. Robust face recognition methods can utilize compact representations extracted from multi-view videos. In the meantime, the constraints of consistent tracking of faces and the limitation of network resources should be satisfied. In this book chapter, we discuss several techniques for continuous tracking of faces acquired by a camera network as well as a face matching algorithm. We conclude this chapter with the discussion on the remaining challenges and emerging frameworks for face recognition in outdoor camera networks.

Learning a classifier from ambiguously labeled face images is challenging since training images are not always explicitly-labeled. For instance, face images of two persons in a news photo are not explicitly labeled by their names in the caption. We propose a Matrix Completion for Ambiguity Resolution (MCar) method for predicting the actual labels from ambiguously labeled images. This step is followed by learning a standard supervised classifier from the disambiguated labels to classify new images. To prevent the majority labels from dominating the result of MCar, we generalize MCar to a weighted MCar (WMCar) that handles label imbalance. Since WMCar outputs a soft labeling vector of reduced ambiguity for each instance, we can iteratively refine it by feeding it as the input to WMCar. Nevertheless, such an iterative implementation can be affected by the noisy soft labeling vectors, and thus the performance may degrade. Our proposed Iterative Candidate Elimination (ICE) procedure makes the iterative ambiguity resolution possible by gradually eliminating a portion of least likely candidates in ambiguously labeled faces. We further extend MCar to incorporate the labeling constraints among instances when such prior knowledge is available. Compared to existing methods, our approach demonstrates improvements on several ambiguously labeled datasets.

Defocus blur usually causes performance degradation in establishing the visual correspondence between stereo images. We propose a blur-aware disparity estimation method that is robust to the mismatch of focus in stereo images. The relative blur resulting from the mismatch of focus between stereo images is approximated as the difference of the square diameters of the blur kernels. Based on the defocus and stereo model, we propose the relative blur versus disparity (RBD) model that characterizes the relative blur as a second-order polynomial function of disparity. Our method alternates between RBD model update and disparity update in each iteration. The RBD model in return refines the disparity estimation by updating the matching cost and aggregation weight to compensate the mismatch of focus. Experiments using both synthesized and real datasets demonstrate the effectiveness of our proposed algorithm.

In real applications, data is not always explicitly-labeled. For instance, label ambiguity exists when we associate two persons appearing in a news photo with two names provided in the caption. We propose a matrix completion-based method for predicting the actual labels from the ambiguously labeled instances, and a standard supervised classifier can learn from the disambiguated labels to classify new data. We further generalize the method to handle the labeling constraints between instances when such prior knowledge is available. Compared to existing methods, our approach achieves 2.9% improvement on the labeling accuracy of the Lost dataset and comparable performance on the Labeled Yahoo! News dataset.

We propose a non-local cost aggregation algorithm to recognize the identity of face and person tracks in a TV-series. In our approach, the fundamental element for identification is a track node, which is built on top of face and person tracks. Track nodes with temporal dependency are grouped into a knot. These knots then serve as the basic units in the construction of a k-knot graph for exploring the video structure. We build the minimum-distance spanning tree (MST) from the k-knot graph such that track nodes of similar appearance are adjacent to each other in MST. Non-local cost aggregation is performed on MST, which ensures information from face and person tracks is utilized as a whole to improve the identification performance. The identification task is performed by minimizing the cost of each knot, which takes into account the unique presence of a subject in a venue. Experimental results demonstrate the effectiveness of our method.

Markerless technology is a game changer for motion-capture applications, such as the monitoring of patients outside the hospital, realistic face-to-face communication across continents, and observation across large spaces.

Representational Change Theory claims that participants form inappropriate representations at the beginning of the insight problem solving process and that these initial representations must be transformed to discover the solution (Knoblich, Ohlsson, Haider, & Rhenius, 1999; Knoblich, Ohlsson, & Raney, 2001; Ohlsson, 1992). The theory also claims that all participants are trapped by inappropriate representations, regardless of the result, but it is easier for successful participants to transform their initial representations. However, the transformation of representations is not the only critical factor. This study investigates the hypothesis that the process of fixedness averting plays an important role in insight problem solving and is helpful for representational change. To verify the influence of fixedness averting on representational change processes, matchstick arithmetic problems were employed as an experimental model. In experiment 1, insight problem solving results could be predicted within the first third of the duration of the task. The gaze duration in the fixation region of successful participants was shorter than the gaze duration of unsuccessful participants. In experiment 2, participants' foci of attention were experimentally manipulated by presenting different animated diagrams to guide their attention. We found that the rate of correct responses was significantly reduced when participants' attention was guided to the fixation region. Representational Change Theory declares that changing inappropriate initial representations is necessary for solving insight problems. The present study demonstrates that in addition to representational change, fixedness averting is also crucial to problem solving. (C) 2014 Elsevier Ltd. All rights reserved.