Shinji Tsuji

Understanding decision-making processes is informative for strategic planning. Aiming to understand human risk-taking behavior in decision-making, we investigate the possibility of classifying whether a shot is offensive or not, targeting table tennis videos. We define the problem in a multi-task setting: detecting shots with frame-level precision while classifying shot offensiveness, and, as an optional task, predicting which player made the shot. We use commercial table tennis videos for target analysis and propose audio-visual self-supervised training, leveraging web videos with similar camera views. Our local contrastive loss encourages the model to learn frame-wise action locality collaboratively with traditional segment-wise contrastive loss, which we call global contrastive loss. Experimental results proved that the collaboration of two contrastive losses boosts the prediction performance.

In Japan, China, andKorea, Bombus ignitus plays a vital role as a pollinator insect in greenhouse horticulture. Over time, the activity of worker bees diminishes, prompting the replacement of the entire hive with new bees. However, considering the cost of hives, it is imperative to pinpoint the optimal replacement timing, balancing the decline in bee activity and the expenses associated with hive renewal. To achieve this objective, calculating the ideal timing for replacement using activity data is preferable. Currently, due to the unavailability of activity data, farmers often rely on visual monitoring and empirical judgments to decide when to replace hives. To address this gap, our study focuses on accurately quantifying the arrivals and departures of Bombus ignitus males and workers from hives, using these metrics as reliable indicators of bee activity. For this purpose, we propose a method for accurately tracking Bombus ignitus . Our proposed method combines specialized tracking techniques for honeybees, drawn from existing research, with commonly used tracking methods for pedestrians and automobiles. The evaluation of our method using test data showcases superior tracking precision with reduced errors, providing a more accurate tally of arrivals and departures compared to existing approaches. Furthermore, when applied on actual farms, our method revealed a significant decrease in worker bee arrivals and departures as the expected replacement time, based on farmers’ experience, drew nearer. This promising result suggests that our approach facilitates data-informed decision-making. Consequently, we anticipate more efficient and appropriately timed hive replacements, based on evidence, in the future.