Hossen Asiful Mustafa

Hossen Asiful Mustafa
Bangladesh University of Engineering and Technology (BUET)

Robust Pose-Based Multi-view Gait Recognition using Recurrent Neural Network

Abstract: Recognizing individual people from gait is still a challenging problem in computer vision research due to the presence of various covariate factors like varying view angle, change in clothing, walking speed, load carriage, etc. Most of the earlier works are based on human silhouettes which have proven to be efficient in recognition but are not invariant to change in illumination and clothing. In this research, to address this problem, we present a simple yet effective approach for robust gait recognition using a recurrent neural network (RNN). Our network with GRU architecture is very powerful in capturing the temporal dynamics of the human body pose sequence and performing recognition. The experimental results on challenging CASIA A and CASIA B gait datasets demonstrate that the proposed method has achieved state-of-the-art performance on both single-view and cross-view gait recognition which proves the effectiveness of our method.

Mohammad Arif Hossain

Mohammad Arif Hossain
JCR Pharmaceuticals Ltd.

Generation and Characterization of Motor Neuron Progenitors and Motor Neurons Using Metachromatic Leukodystrophy-Induced Pluripotent Stem Cells

Abstract: The pathological consequences of primary storage, autophagy impairment, impaired mitochondrial dynamics, and endoplasmic reticulum (ER) stress on neural cell dysfunction and apoptosis in metachromatic leukodystrophy (MLD) have been poorly elucidated. In the present study, we generated two cell lines of patient-specific-induced pluripotent stem cells (iPSCs) and modeled the progression of pathological events during the differentiation of iPSCs to motor neuron progenitors (MNPs) and mature motor neurons (MNs). The iPS cells were generated from two late-infantile MLD patient-derived skin fibroblasts using electroporation or the Sendai virus. Olig2+ MNPs were generated from both iPSC lines using a combination of small molecules in a chemically defined neural medium. Furthermore, the MNPs could be differentiated into mature MNs, confirmed by RT–PCR and MN markers, including SMI32 and ChAT. The population of MNs was approximately 50% under cultural conditions. Pathological observation of MLD patient-derived iPSCs revealed lysosomal accumulation and impaired autophagy. In addition, both MNPs and MNs derived from MLD-iPSCs showed increased lysosomal collection, dysfunctional autophagy, impaired mitophagy, endoplasmic reticulum (ER) stress, or unfolded protein response (UPR) activation, and premature cellular death.

Manabu Sugimoto

Manabu Sugimoto
Kumamoto University

Electronic-Structure Informatics: Development and Challenges

Abstract: Molecular properties reflect their structural features. This fact tempts us to establish the structure-property relationship, which is expected to be practically very useful for discovery and optimization of materials. Irrespective of the usefulness, the relationship is superficial because both molecular structure and properties are dependent on electronic structure of a molecule. From this viewpoint, we have been suggesting “Electronic-Structure Informatics (ESI)” in which information related to the electronic structure is numerically quantified for machine learning. In this talk, we will introduce our ESI approach and the recent progress in its application. Some future challenges in ESI and the related field will also be discussed.

Katsutoshi Masai

Katsutoshi Masai
NTT Communication Science Laboratories; Keio University

Optical sensor-based smart eyewear for facial expression recognition

Abstract: This talk will introduce smart eyewear that can recognize facial expressions in daily life. Facial expressions convey important information about a person's inner world. Although there are various automatic recognition technologies available, there have been no methods suitable for measuring facial expressions in daily life due to the burden of wearing comfort and limitations in the range of use and so on. To address these issues, we propose a facial expression recognition method based on optical sensors. This method can estimate subtle facial expressions. This talk will also go over its applications to HMDs as well as facial and eye gesture recognition techniques.

Wataru Sato

Wataru Sato
RIKEN

Emotion dynamics sensing using a wearable facial EMG device

Abstract: Emotion dynamics sensing using physiological signals in real-life situations can be practically valuable. However, no study determined whether emotional valence dynamics can be estimated, specifically using wearable devices. To investigate this issue, we conducted two experiments. In Study 1, we measured participants' facial EMG from the corrugator supercilii and zygomatic major muscles while viewing emotional films and assessed their cued-recall continuous valence ratings. The corrugator and zygomatic EMG were negatively and positively associated with valence ratings, respectively. In Study 2, we developed a wearable device to record facial EMG and measured participants' facial EMG while playing Wii Bowling games and their cued-recall continuous valence ratings. Facial EMG activities were associated with the valence ratings. These data suggest that facial EMG signals recorded by a wearable device can be used to assess emotional valence dynamics in real-life situations.

Kazunori Terada

Kazunori Terada
Gifu University

Inferring Mental States from Emotion Expressions and Social Decision Making

Abstract: Emotions have social functions, such as dominating and rejecting competitive opponents by showing anger and maintaining a long-term reciprocal relationship with cooperative partners by presenting joy or friendship. Underlying social cognitive mechanism to realize this function is called a reverse appraisal process, in which the observer not only reacts reflectively to the counterpart's emotional expressions but also infers the mental states of the counterpart and outputs appropriate actions. The mental state inference is realized by Bayesian inference using a generative model (appraisal model). In this presentation, our studies on human social decision-making by reading the mental states of others through interaction with AI agents will be introduced.

Mutsumi Kimura

Mutsumi Kimura
Ryukoku University

Neuromorphic Systems using Novel Memdevices

Abstract: We are investigating neuromorphic systems using novel memdevices, such as, memristors and memcapacitors, with novel materials, such as, amorphous and poly-crystal metal-oxide semiconductors. I will give a talk on the theory, principle, performance, etc. of them in details. These systems and materials will be available in many cases in the research fields of Data Science.

Michihiro Shintani

Michihiro Shintani
Kyoto Institute of Technology

Wafer-Level Variation Modeling for Efficient LSI Testing

Abstract: In the manufacture of large-scale integrated circuits (LSIs), testing plays an important role to ensure their reliability. On the other hand, test costs account for the majority of manufacturing costs due to the increasing miniaturization and complexity of LSIs. Within this context, statistical wafer-level variation modeling is an attractive method for reducing the measurement cost in LSI testing while maintaining the test quality. In this presentation, statistical wafer-level modeling methods will be introduced, which were recently presented by our group.

Gustavo Garcia

Gustavo Garcia
Ritsumeikan University

Robotics Technologies and their Benchmarking at International Robotics Competitions

Abstract: The talk is about the system development approach followed by researchers and engineers of the team NAIST-RITS-Panasonic (Japan) for their participation in international robotics competitions, which are organized to foster innovation and to bring together the research community to solve real-world, current problems such as drilling in aircraft manufacturing (Airbus Shopfloor Challenge), warehouse automation (Amazon Robotics Challenge), and, most recently, convenience store automation (Future Convenience Store Challenge). I will talk about how the team built different robots with multiple end effectors, as well as different technologies for mobile manipulation, vision, and HRI.