Facial expression recognition (FER) is a challenging problem in the fields of pattern recognition and computer vision. The recent success of convolutional neural networks (CNNs) in object detection and object segmentation tasks has shown promise in building an automatic deep CNN-based FER model. However, in real-world scenarios, performance degrades dramatically owing to the great diversity of factors unrelated to facial expressions, and due to a lack of training data and an intrinsic imbalance in the existing facial emotion datasets. To tackle these problems, this paper not only applies deep transfer learning techniques, but also proposes a novel loss function called weighted-cluster loss, which is used during the fine-tuning phase. Specifically, the weighted-cluster loss function simultaneously improves the intra-class compactness and the inter-class separability by learning a class center for each emotion class. It also takes the imbalance in a facial expression dataset into account by giving each emotion class a weight based on its proportion of the total number of images. In addition, a recent, successful deep CNN architecture, pre-trained in the task of face identification with the VGGFace2 database from the Visual Geometry Group at Oxford University, is employed and fine-tuned using the proposed loss function to recognize eight basic facial emotions from the AffectNet database of facial expression, valence, and arousal computing in the wild. Experiments on an AffectNet real-world facial dataset demonstrate that our method outperforms the baseline CNN models that use either weighted-softmax loss or center loss.

No abstract

Location prediction plays an important role in modeling human mobility. Existing studies focused on developing a prediction model which is based solely on the past mobility of only the person of interest (POI), rather than including information on the mobility of her/his companions. In fact, people frequently move in a group, and thus, using mobility data of a person's companions can enhance accuracy when predicting that person's future locations. Motivated by this, we propose a two-phase framework for predicting an individual's future locations that fully benefits from spatio-temporal contexts embedded in that person's and his/her companions' mobility. The framework first determines the POI's companions, then predicts future locations based on mobility information for both the POI and selected companions. Two companion selection methods are proposed in this work. The first method uses spatial closeness (SC) to determine the companions of the POI by measuring the similarity of the individuals' geographic distributions. The second method builds person ID embedding (PIE) vectors, and cosine similarity is used to select the POI's companions. To mitigate the curse of dimensionality, the framework also uses a stacked autoencoder in which the encoder compresses a high-dimensional input feature (e.g., location, time, and person ID) into a low-dimensional latent vector. For the second phase of the framework, a bidirectional recurrent neural network (BRNN)-based multi-output model is proposed to predict a person's future locations in the next several time slots. To train the BRNN model, weighted loss is used, which takes into account the importance of each future time slot to predict the POI's locations accurately. Experiments are conducted on two largescale Wi-Fi trace datasets, demonstrating that the proposed model can effectively predict human future locations.

No abstract

No abstract