For a long time, anomaly localization has been widely used in industries. Previous studies focused on approximating the distribution of normal features without adaptation to a target dataset. However, since anomaly localization should precisely discriminate between normal and abnormal features, the absence of adaptation may make the normality of abnormal features overestimated. Thus, we propose Coupled-hypersphere-based Feature Adaptation (CFA) which accomplishes sophisticated anomaly localization using features adapted to the target dataset. CFA consists of (1) a learnable patch descriptor that learns and embeds target-oriented features and (2) a scalable memory bank independent of the size of the target dataset. And, CFA adopts transfer learning to increase the normal feature density so that abnormal features can be clearly distinguished by applying patch descriptor and memory bank to a pre-trained CNN. The proposed method outperforms the previous methods quantitatively and qualitatively. For example, it provides an AUROC score of 99.5% in anomaly detection and 98.5% in anomaly localization of MVTec AD benchmark. In addition, this paper points out the negative effects of biased features of pre-trained CNNs and emphasizes the importance of adaptation to the target dataset.

Facial emotion recognition (FER) has been an active research topic in the past several years. One of difficulties in FER is the effective capture of geometrical and temporary information from landmarks. In this paper, we propose a graph convolution neural network that utilizes landmark features for FER, which we called a directed graph neural network (DGNN). Nodes in the graph structure were defined by landmarks, and edges in the directed graph were built by the Delaunay method. By using graph neural networks, we could capture emotional information through faces’ inherent properties, like geometrical and temporary information. Also, in order to prevent the vanishing gradient problem, we further utilized a stable form of a temporal block in the graph framework. Our experimental results proved the effectiveness of the proposed method for datasets such as CK+ (96.02%), MMI (69.4%), and AFEW (32.64%). Also, a fusion network using image information as well as landmarks, is presented and investigated for the CK+ (98.47% performance) and AFEW (50.65% performance) datasets.

Emotion recognition based on facial expressions is very important for effective interaction of humans with artificial intelligence (AI) systems such as social robots. On the other hand, in real environment, it is much harder to recognize facial micro-expressions (FMEs) than facial general-expressions having rich emotions. In this paper, we propose a two-dimensional (2D) landmark feature map for effectively recognizing such FMEs. The proposed 2D landmark feature map (LFM) is obtained by transforming conventional coordinate-based landmark information into 2D image information. LFM is designed to have an advantageous property independent of the intensity of facial expression change. Also, we propose an LFM-based emotion recognition method that is an integrated framework of convolutional neural network (CNN) and long short-term memory (LSTM). Experimental results show that the proposed method achieves about 71% and 74% in the well-known micro-expression datasets, i.e., SMIC and CASME II, respectively, which outperforms the conventional methods. The performance of the proposed method was also verified through experiments on composite micro-expression dataset, which consists of SMIC, CAMSE II and SAMM, and cross-dataset validation using SMIC and CAMSE II. In addition, we prove that the proposed method is independent of facial expression intensity through an experiment on CK+ dataset. Finally, we demonstrate that the proposed method is valid even for the MAHNOB-HCI and MEVIEW datasets that are produced to monitor actual and wild emotional responses.

This paper proposes various techniques that help Vision Transformer (ViT) to learn small-size datasets from scratch successfully. ViT, which applied the transformer structure to the image classification task, has outperformed convolutional neural networks, recently. However, the high performance of ViT results from pre-training using large-size datasets, and its dependence on large datasets comes from low locality inductive bias. And conventional ViT cannot effectively attend the target class due to redundant attention caused by a rather high constant temperature factor. In order to improve the locality inductive bias of ViT, this paper proposes novel tokenization (Shifted Patch Tokenization: SPT) using shifted patches and a position encoding (CoordConv Position Encoding: CPE) using 1×1 CoordConv. Also, to improve poor attention, we propose a new self-attention mechanism (Locality Self-Attention: LSA) based on learnable temperature and self-relation masking. SPT, CPE, and LSA are intuitive techniques, but they successfully improve the performance of ViT even on small-size datasets. We qualitatively show that each technique attends a more important area and contributes to having a flatter loss landscape. Moreover, the proposed techniques are generic add-on modules applicable to various ViT backbones. Our experiments show, when learning Tiny-ImageNet from scratch, the proposed scheme based on SPT, CPE, and LSA increases the accuracy of ViT backbones by +3.66 on average and up to +5.7. Also, the performance improvement of ViT backbones in ImageNet-1K classification, learning on COCO from scratch, and transfer learning on classification datasets verify that the generalization ability of the proposed method is excellent.

Emotion recognition is a very important technique for ultimate interactions between human beings and artificial intelligence systems. For effective emotion recognition in a continuous-time domain, this article presents a multimodal fusion network which integrates video modality and electroencephalogram (EEG) modality networks. To calculate the attention weights of facial video features and the corresponding EEG features in fusion, a multimodal attention network, that is utilizing bilinear pooling based on low-rank decomposition, is proposed. Finally, continuous domain valence values are computed by using two modality network outputs and attention weights. Experimental results show that the proposed fusion network provides an improved performance of about 6.9% over the video modality network for the MAHNOB human computer interface (MAHNOB-HCI) dataset. Also, we achieved the performance improvement even for our proprietary dataset.