Stephanie Nix

This paper presents a novel approach for predicting Particulate Matter (PM) concentrations using mobile camera devices. In response to persistent air pollution challenges across Japan, we developed a system that utilizes cutting-edge transformer-based deep learning architectures to estimate PM values from imagery captured by smartphone cameras. Our approach employs Contrastive Language–Image Pre-Training (CLIP) as a multimodal framework to extract visual features associated with PM concentration from environmental scenes. We first developed a baseline through comparative analysis of time-series models for 1D PM signal prediction, finding that linear models, particularly NLinear, outperformed complex transformer architectures for short-term forecasting tasks. Building on these insights, we implemented a CLIP-based system for 2D image analysis that achieved a Top-1 accuracy of 0.24 and a Top-5 accuracy of 0.52 when tested on diverse smartphone-captured images. The performance evaluations on Graphics Processing Unit (GPU) and Single-Board Computer (SBC) platforms highlight a viable path toward edge deployment. Processing times of 0.29 s per image on the GPU versus 2.68 s on the SBC demonstrate the potential for scalable, real-time environmental monitoring. We consider that this research connects high-performance computing with energy-efficient hardware solutions, creating a practical framework for distributed environmental monitoring that reduces reliance on costly centralized monitoring systems. Our findings indicate that transformer-based multimodal models present a promising approach for mobile sensing applications, with opportunities for further improvement through seasonal data expansion and architectural refinements.

This study proposes an object detector for apple trees as a first step in developing agricultural digital twins. An original dataset of orchard images was created and used to train Single Shot MultiBox Detector (SSD) and You Only Look Once (YOLO) models. Performance was evaluated using mean Average Precision (mAP). YOLO significantly outperformed SSD, achieving 91.3% mAP compared to the SSD’s 46.7%. Results indicate YOLO’s Darknet-53 backbone extracts more complex features suited to tree detection. This work demonstrates the potential of deep learning for automated data collection in smart farming applications.

Understanding the transport of small molecules such as oxygen in biological systems requires knowledge about how molecules dynamically interact with these molecules. This study investigates how red blood cells influence the diffusion of small molecules in simple shear flow by coupling Brownian dynamics simulations with a finite element–boundary integral method to simulate particle transport near spherical and red blood cells. The simulation found that the presence of a rotating cell significantly reduces the effective diffusion rate of small molecules. Specifically, the circulatory flow induced by cell deformation during tank-treading motion leads to a diminished mean squared displacement of particles. Notably, a tumbling red blood cell produces a more pronounced effect on particle motion compared with a spherical cell under identical flow conditions. This research has broader implications for understanding complex diffusion processes in various biological systems, by highlighting the complex interactions between cellular motion and molecular transport.

This study is intended to develop a stress measurement and visualization system for stress management in terms of simplicity and reliability. We present a classification and visualization method of mood states based on unsupervised machine learning (ML) algorithms. Our proposed method attempts to examine the relation between mood states and extracted categories in human communication from facial expressions, gaze distribution area and density, and rapid eye movements, defined as saccades. Using a psychological check sheet and a communication video with an interlocutor, an original benchmark dataset was obtained from 20 subjects (10 male, 10 female) in their 20s for four or eight weeks at weekly intervals. We used a Profile of Mood States Second edition (POMS2) psychological check sheet to extract total mood disturbance (TMD) and friendliness (F). These two indicators were classified into five categories using self-organizing maps (SOM) and U-Matrix. The relation between gaze and facial expressions was analyzed from the extracted five categories. Data from subjects in the positive categories were found to have a positive correlation with the concentrated distributions of gaze and saccades. Regarding facial expressions, the subjects showed a constant expression time of intentional smiles. By contrast, subjects in negative categories experienced a time difference in intentional smiles. Moreover, three comparative experiment results demonstrated that the feature addition of gaze and facial expressions to TMD and F clarified category boundaries obtained from U-Matrix. We verify that the use of SOM and its two variants is the best combination for the visualization of mood states.

Classification, segmentation, and recognition techniques based on deep-learning algorithms are used for smart farming. It is an important and challenging task to reduce the time, burden, and cost of annotation procedures for collected datasets from fields and crops that are changing in a wide variety of ways according to growing, weather patterns, and seasons. This study was conducted to generate crop image datasets for semantic segmentation based on an image style transfer using generative adversarial networks (GANs). To assess data-augmented performance and calculation burdens, our proposed framework comprises contrastive unpaired translation (CUT) for a conditional GAN, pix2pixHD for an unconditional GAN, and DeepLabV3+ for semantic segmentation. Using these networks, the proposed framework provides not only image generation for data augmentation, but also automatic labeling based on distinctive feature learning among domains. The Fréchet inception distance (FID) and mean intersection over union (mIoU) were used, respectively, as evaluation metrics for GANs and semantic segmentation. We used a public benchmark dataset and two original benchmark datasets to evaluate our framework of four image-augmentation types compared with the baseline without using GANs. The experimentally obtained results showed the efficacy of using augmented images, which we evaluated using FID and mIoU. The mIoU scores for the public benchmark dataset improved by 0.03 for the training subset, while remaining similar on the test subset. For the first original benchmark dataset, the mIoU scores improved by 0.01 for the test subset, while they dropped by 0.03 for the training subset. Finally, the mIoU scores for the second original benchmark dataset improved by 0.18 for the training subset and 0.03 for the test subset.

This paper presents an estimation method for a sound source of pre-recorded mallard calls from acoustic information using two microphone arrays combined with delay-and-sum beamforming. Rice farming using mallards saves labor because mallards work instead of farmers. Nevertheless, the number of mallards declines when they are preyed upon by natural enemies such as crows, kites, and weasels. We consider that efficient management can be achieved by locating and identifying the locations of mallards and their natural enemies using acoustic information that can be widely sensed in a paddy field. For this study, we developed a prototype system that comprises two sets of microphone arrays. We used 64 microphones in all installed on our originally designed and assembled sensor mounts. We obtained three acoustic datasets in an outdoor environment for our benchmark evaluation. The experimentally obtained results demonstrated that the proposed system provides adequate accuracy for application to rice–duck farming.