Automated Behavior Analysis Using a YOLO-Based Object Detection System

Amino, Kai; Matsuo, Takashi

doi:10.1007/978-1-0716-2321-3_14

Cited by 6 publications

(3 citation statements)

References 32 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…Videos were analyzed using a custom-made YOLO-based object detection system (Amino & Matsuo, 2022), which can identify the males marked with each color, as well as the occurrence of boxing behavior. For each contest, we calculated the total time spent by each male on the food podium.…”

Section: Discussionmentioning

confidence: 99%

Effects of a past contest on the future winning probability in a hyper‐aggressive fruit fly

Amino

Matsuo

2023

Ethology

Self Cite

View full text Add to dashboard Cite

In various animal taxa, prior experience of contests influences the consequences of subsequent contests. A well‐known example is the winner and loser effects, in which winners have a higher winning probability, whereas losers become more likely to lose in the subsequent contests. Previous studies have suggested that the intensity of past contests also affects the future winning probability in two ways: modulating the extent of the winner–loser effect or decreasing the winning probability of both winners and losers. However, the number of studies that investigated the effect of both the contest outcome and intensity has been limited. Here, we analyzed the effect of past fighting experience in a hyper‐aggressive fruit fly Drosophila prolongata to ask (1) whether the winner–loser effect exists in this species and (2) how the contest intensity influences the subsequent winning probability. To disentangle the multiple factors affecting the winning probability, we employed the experimental design involving both body size manipulation and the comparison of different intervals between the two contests. The results did not support the existence of the winner–loser effect: both winners and losers of the first contest showed decreased winning probability in the second contest. With a short interval (2 h), contest intensity negatively influenced the winning probability in winners, though the significance level of its effect was marginally above the statistical threshold. In contrast, contest intensity had no effect in losers. Notably, contest intensity positively correlated with the winning probability after the longer interval (1 d) in both winners and losers, suggesting that it reflects the intrinsic fighting potential of individuals. The possible mechanisms underlying the observed results were discussed. Our study illustrates the transient influence of past contests other than the winner–loser effect.

show abstract

Section: Discussionmentioning

confidence: 99%

Effects of a past contest on the future winning probability in a hyper‐aggressive fruit fly

Amino

Matsuo

2023

Ethology

Self Cite

View full text Add to dashboard Cite

show abstract

“…RFID systems also play an important role in invertebrate conservation efforts (Kissling et al, 2014; Barlow et al, 2019). The use of object detection and marker recognition algorithms to study invertebrate behaviour is still in its infancy, though recent papers have used such algorithms to behaviour in lab and occasionally wild settings (Hatton-Jones et al, 2021; Petso et al, 2021; Amino and Matsuo, 2022). However, such algorithms have received wider use in studies of vertebrate behaviour and in species identification for camera trapping (Manning et al, 2019; Schindler & Steinhage 2021; Petso et al, 2021).…”

Section: Introductionmentioning

confidence: 99%

Measuring the effect of RFID and Marker Recognition tags on cockroach behaviour using AI aided tracking

McLean,

Fisher

2024

Preprint

View full text Add to dashboard Cite

RFID technology and marker recognition algorithms can offer an efficient and non-intrusive means of tracking animal positions. As such, they have become important tools for invertebrate behavioural research. Both approaches require fixing a tag or marker to the study organism, and so it is useful to quantify the effects such procedures have on behaviour before proceeding with further research. However, frequently studies do not report doing such tests. Here, we demonstrate a time-efficient and accessible method for quantifying the impact of tagging on individual movement using open-source automated video tracking software. We tested the effect of RFID tags and tags suitable for marker recognition algorithms on the movement of Argentinian wood roaches (Blapicta dubia) by filming tagged and untagged roaches in laboratory conditions. We employed DeepLabCut on the resultant videos to track cockroach movement and extract measures of behavioural traits. We found no statistically significant differences between RFID tagged and untagged groups in average speed over the trial period, the number of unique zones explored, and the number of discrete walks. However, groups that were tagged with labels for marker recognition had significantly higher values for all three metrics. We therefore support the use of RFID tags to monitor the behaviour ofB. dubiabut note that the effect of using labels suitable for label recognition to identify individuals should be taken into consideration when measuringB.dubiabehaviour. We hope that this study can provide an accessible and viable roadmap for further work investigating the effects of tagging on insect behaviour.

show abstract

“…Object detection, as a core component of computer vision, facilitates the recognition and localization of entities within digital images or videos. Its applications are vast, ranging from enhancing surveillance capabilities to enabling safe navigation in autonomous vehicles [1], [2], [3], [4], [5], [6], [7], [8], [9]. However, the efficacy of object detection is markedly challenged in scenarios characterized by low-light conditions, where visual information is limited.…”

Section: Introductionmentioning

confidence: 99%

A Comparative Study of YOLOv8 and YOLO - NAS Performance in Human Detection Image

Hendrawan,

Kolandaisamy

2023

JTMI

View full text Add to dashboard Cite

In the realm of computer vision, object detection holds immense importance across applications such as surveillance and autonomous vehicles. This study addresses the critical challenge of human detection under low-light conditions, essential for nocturnal surveillance and autonomous driving systems. Focusing on the evolution of YOLO models, particularly YOLO - NAS and YOLOv8, a research gap is identified concerning their performance in low-light scenarios. The research conducts a detailed analysis of YOLO - NAS and YOLOv8 effectiveness in human detection under reduced ambient illumination. Object detection, vital in computer vision, faces challenges in low-light scenarios. This study concentrates on human detection due to its significance in night-time surveillance and autonomous driving. Despite YOLO models' evolution, a research gap exists in comparing their performance in low-light conditions. The study aims to fill this gap, providing insights for enhancing human detection methodologies in challenging lighting environments.

show abstract

Automated Behavior Analysis Using a YOLO-Based Object Detection System

Cited by 6 publications

References 32 publications

Effects of a past contest on the future winning probability in a hyper‐aggressive fruit fly

Effects of a past contest on the future winning probability in a hyper‐aggressive fruit fly

Measuring the effect of RFID and Marker Recognition tags on cockroach behaviour using AI aided tracking

A Comparative Study of YOLOv8 and YOLO - NAS Performance in Human Detection Image

Contact Info

Product

Resources

About