A5036735172- Gaze Tracking and Assistive Technology
- Visual Attention and Saliency Detection
- Human-Automation Interaction and Safety
- Advanced Image and Video Retrieval Techniques
- Autonomous Vehicle Technology and Safety
- Video Surveillance and Tracking Methods
- Video Analysis and Summarization
- Human Pose and Action Recognition
- Traffic and Road Safety
- Face recognition and analysis
- Face and Expression Recognition
- Image Retrieval and Classification Techniques
- Robotics and Sensor-Based Localization
- Multimodal Machine Learning Applications
- Advanced Neural Network Applications
- Safety Warnings and Signage
- Color perception and design
- Advanced Vision and Imaging
- Image and Video Quality Assessment
- Gait Recognition and Analysis
- Speech and dialogue systems
- Hand Gesture Recognition Systems
- Anomaly Detection Techniques and Applications
- Food Quality and Safety Studies
- Biometric Identification and Security
Nagoya University
2015-2024
University of Human Environments
2021-2024
Kansai Electric Power (Japan)
2021
Television Research Institute
2021
University of Crete
2021
FORTH Institute of Computer Science
2021
Foundation for Research and Technology Hellas
2021
Institute of Electrical and Electronics Engineers
2019
Chukyo University
2018
Kyoto University
2005-2012
We humans are easily able to instantaneously detect the regions in a visual scene that most likely contain something of interest. Exploiting this pre-selection mechanism called attention for image and video processing systems would make them more sophisticated therefore useful. This paper briefly describes various computational models human their development, as well related psychophysical findings. In particular, our objective is carefully distinguish several types studies saliency measure...
Compared to a manual driving vehicle (MV), an automated lacks way communicate with the pedestrian through driver when it interacts because usually does not participate in tasks. Thus, external human machine interface (eHMI) can be viewed as novel explicit communication method for providing intentions of (AV) pedestrians they need negotiate interaction, e.g., encountering scene. However, eHMI may guarantee that will fully recognize intention AV. In this paper, we propose instruction eHMI's...
We investigated a method to estimate the degree of driver awareness surrounding vehicles based on correlation between gaze direction and risks caused by vehicles. The posed were represented their time collision (TTC) from driver's vehicle. recorded driving data five expert non-expert drivers while passing other expressways, using an instrumented manually labeled drivers' directions video faces, detected positions calculated TTC laser scanners mounted front back focused behavior for seconds...
This paper proposes a hybrid localization method that fuses Monte Carlo (MCL) and convolutional neural network (CNN)-based end-to-end (E2E) localization. MCL is based on particle filter requires proposal distributions to sample the particles. The distribution generally predicted using motion model. However, because model cannot handle unanticipated errors, sometimes inaccurate. use of other ideal distributions, such as measurement model, can improve robustness against errors. technique...
Small Object Detection (SOD) is an important machine vision topic because (i) a variety of real-world applications require object detection for distant objects and (ii) SOD challenging task due to the noisy, blurred, less-informative image appearances small objects. This paper proposes new dataset consisting 39,070 images including 137,121 bird instances, which called Spotting Birds (SOD4SB) dataset. The detail challenge with SOD4SB <sup xmlns:mml="http://www.w3.org/1998/Math/MathML"...
To support safe driving, numerous methods of detecting distractions using measurements a driver's gaze have been proposed. These empirically focused on certain driving contexts and analyzed behavior under particular peripheral vehicle conditions; therefore, situations were not considered. address this problem with hypothesis-testing approaches, we turn the around propose data-mining approach that analyzes during transitions drivers in order to compare their neutral state cognitive...
Semantic segmentation is an interesting task for many deep learning researchers scene understanding. However, recognizing details about objects' attributes can be more informative and also helpful a better understanding in intelligent vehicle use cases. This paper introduces method simultaneous semantic pedestrian recognition. A modified dataset built on top of the Cityscapes created by adding attribute classes corresponding to orientation attributes. The proposed extends SegNet model...
Interactions between pedestrians and automated vehicles (AVs) will increase significantly with the popularity of AV. However, often have not enough trust on AVs , particularly when they are confused about an AV's intention in a interaction. This study seeks to evaluate if clearly understand driving intentions interactions presents experimental research relationship gaze behaviors their understanding The hypothesis investigated this was that less pedestrian understands AV, longer duration...
People are being inundated under enormous volumes of information and they often dither about making the right choices from these. Interactive user support by service system such as concierge services will effectively assist people. However, human-machine interaction still lacks naturalness thoughtfulness despite widespread utilization intelligent systems. The needs to estimate user's interest improve choices. We propose a novel approach estimating interest, which is based on relationship...
This paper presents a comparison of driving behavior modeling methods based on hidden Markov models (HMMs) with driver's eye-gaze measurement and ego-vehicle localization. Original HMMs are sometimes insufficient to model real-world scenarios. To overcome these limitations, extended have been proposed, e.g., autoregressive input-output (AIOHMMs). first details AIOHMMs ways use them for modeling. We compare the performance maneuver discrimination six types HMMs. The data this work was...
Image captioning can show great performance for generating captions general purposes, but it remains difficult to adjust the generated different applications. In this paper, we propose an image method which generate both imageability- and length-controllable captions. The imageability parameter adjusts level of visual descriptiveness caption, making either more abstract or concrete. contrast, length only caption while keeping on a similar degree. Based transformer architecture, our model is...
In this paper, we investigate a method for detecting risky lane changes using integrated modeling of driver gaze and vehicle operation behavior. Driver direction behavior are broken down into discrete acts, e.g., looking in the rear view mirror, braking, etc., sequences these actions jointly modeled multi-stream hidden Markov models (HMMs). Driving data is recorded on expressways as drivers pass leading vehicles, i.e., make two changes, first to vehicles then move back their original lanes....
Recognizing misalignment between sensor measurements and objects that exist on a map due to inaccuracies in localization estimation is challenging. This can be attributed the fact are individually modeled for solving problem, resulting entire relations of being ignored. letter proposes recognition method using Markov random fields with fully connected latent variables detection failures. The proposed estimates classes each measurement aligned, misaligned, obtained from unknown obstacles....
We investigate a possible method for detecting driver's negative adaptation to an automated driving system by analyzing consistency of driver decision making and gaze behavior during driving. focus on equivalent Level 2 automation per the NHTSA's definition. At this level automation, drivers must be ready take control vehicle in critical situations monitoring environment behavior. Since are not required operate pedals or steering wheel driving, needs detected from other than operation. In...
Semantics can be leveraged in ego-vehicle localization to improve robustness and accuracy because objects with the same labels correctly matched each other. Object recognition has significantly improved owing advances machine learning algorithms. However, perfect object is still challenging real environments. Hence, uncertainty of must considered localization. This letter proposes a novel method that integrates supervised method, which predicts probabilistic distributions over classes for...
Although automated driving systems have been used frequently, they are still unpopular in society. To increase the popularity of vehicles (AVs), assisting pedestrians to accurately understand intentions and improving their perception safety when interacting with AVs considered effective. Therefore, AV should send information about its intention interact each other. However, following questions be answered regarding how sends them: 1) What timing for an make after being noticed by them? 2)...
This paper presents a LiDAR-based 3D Monte Carlo localization (MCL) with an efficient distance field (DF) representation method. To implement MCL, high computing capacity is required because the likelihood of many pose candidates, i.e., particles, must be calculated in real time by comparing sensor measurements and map. Additionally, large-scale map needed for allocation to embedded computers since autonomous vehicles are navigate wide areas. These make it difficult MCL implementation. first...
Most of the recent automated driving systems assume accurate functioning localization. Unanticipated errors cause localization failures and result in driving. An exact failure detection is necessary to ensure safety driving; however, challenging because sensor measurement assumed be independent each other process. Owing assumption, entire relation ignored. Consequently, it difficult recognize misalignment between map when partial overlaps with map. This paper proposes a method for using...
The spatio-temporal correlation analysis between visual saliency and eye movements is presented for the estimation of mental focus toward videos. We extract dynamics patterns areas from videos, which we refer to as saliency-dynamics patterns, evaluate based on their with in view. Experimental results using TV commercials demonstrate effectiveness proposed method mental-focus estimation.
To robustly localize the pose of an ego vehicle within a dynamic environment, it is important to model sensor measurements precisely, including changes in environment. This study describes observation models developed for localization performed highly environments, and presents results comparing these models. In this study, four models, our previously proposed model, were compared by conducting simulation. The had different ways coping with produced results. Moreover, comparison revealed...
Text-to-Image (T2I) generation is the task of synthesizing images corresponding to a given text input. The recent innovations in artificial intelligence have enhanced capacity conventional T2I generation, yielding more and powerful models day by day. However, their behavior known become unstable face inputs containing nonwords that no definition within language. This not only results situations where image does match human expectations but also hinders these from being utilized...
A car driver’s cognitive distraction is a main factor behind accidents. One’s state of mind subconsciously exposed as reaction reflecting it by external stimuli. visual event that occurs in front the driver when peripheral vehicle overtakes regarded stimulus. We focus on temporal relationships between eye gaze and behavior. The analysis result showed depend state. In particular we confirmed timing toward stimulus under distracted induced music retrieval task using an automatic speech...