Daily activity timing in the Anthropocene

Trends in Ecology & Evolution, Journal Year: 2022, Volume and Issue: 38(4), P. 324 - 336

Published: Nov. 16, 2022

Language: Английский

---

Human vs. machine: Detecting wildlife in camera trap images

Ecological Informatics, Journal Year: 2022, Volume and Issue: 72, P. 101876 - 101876

Published: Oct. 27, 2022

As the capacity to collect and store large amounts of data expands, identifying evaluating strategies efficiently convert raw into meaningful information is increasingly necessary. Across disciplines, this processing task has become a significant challenge, delaying progress actionable insights. In ecology, growing use camera traps (i.e., remotely triggered cameras) on wildlife led an enormous volume images) in need review annotation. To expedite trap image processing, many have turned field artificial intelligence (AI) machine learning models automate tasks such as detecting classifying images. contribute understanding utility AI tools for images, we evaluated performance state-of-the-art computer vision model developed by Microsoft Earth named MegaDetector using from ongoing study Arctic Alaska, USA. Compared labels determined manual human review, found reliably presence or absence images generated motion detection settings (≥94.6% accuracy), however, was substantially poorer collected with time-lapse (≤61.6% accuracy). By examining where failed detect wildlife, gained practical insights animal size distance limits discuss how those may impact other systems. We anticipate our findings will stimulate critical thinking about tradeoffs automated process help inform effective implementation designs.

Language: Английский

---

Beyond observation: Deep learning for animal behavior and ecological conservation

Ecological Informatics, Journal Year: 2024, Volume and Issue: unknown, P. 102893 - 102893

Published: Nov. 1, 2024

Language: Английский

---

Large‐scale mammal monitoring: The potential of a citizen science camera‐trapping project in the United Kingdom

Ecological Solutions and Evidence, Journal Year: 2022, Volume and Issue: 3(4)

Published: Oct. 1, 2022

Abstract In light of global biodiversity loss, there is an increasing need for large‐scale wildlife monitoring. This difficult mammals, since they can be elusive and nocturnal. the United Kingdom, a lack systematic, widespread mammal monitoring, recognized deficiency data. Innovative new approaches are required. We developed MammalWeb, portal to enable UK‐wide camera trapping by network citizen scientists partner organizations. MammalWeb contribute both collection classification trap Following trials in 2013–2017, has grown organically increase its geographic reach (e.g. ∼2000 sites Britain). It so far provided equivalent over 340 trap‐years wild produced nearly 440,000 classified image sequences videos, which, 180,000 detections. describe background, development novel we have participation. consider data collected participants, especially their relevance main goals monitoring: provide spatial data, abundance temporal behavioural complement existing Explicit accounting patterns animal activity enables bias relative ad hoc observational Estimating presents challenges, as many camera‐trapping studies, but discuss potential stand, opportunities advance value estimation. Challenges remain MammalWeb's central missions enhancing engagement with connection nature, delivering policy‐relevant on Britain's mammals. these challenges advances respect engagement, science financial security. Our approach reduces administrative burden increases coverage and, such, provides useful addition case studies program design. believe important step towards fulfilling calls monitoring our description identifies agenda that purpose.

Language: Английский

---

A semi‐automated camera trap distance sampling approach for population density estimation

Remote Sensing in Ecology and Conservation, Journal Year: 2023, Volume and Issue: 10(2), P. 156 - 171

Published: Aug. 28, 2023

Abstract Camera traps have become important tools for the monitoring of animal populations. However, study‐specific estimation detection probabilities is key if unbiased abundance estimates unmarked species are to be obtained. Since this process can very time‐consuming, we developed first semi‐automated workflow animals any size and shape estimate population densities. In order obtain observation distances, a deep learning algorithm used create relative depth images that calibrated with small set reference photos each location, distances then extracted automatically detected by MegaDetector 4.0. Animal was generally independent distance camera trap 10 at two different study sites. If an both manually automatically, difference in often minimal about 4 m from trap. The increased approximately linearly larger distances. Nonetheless, density based on manual sampling workflows did not differ significantly. Our results show readily available software reliably within workflow, reducing time required data processing, >13‐fold. This greatly improves accessibility wildlife research management.

Language: Английский

---

A model of optimal digestive strategy in infrequently-feeding snakes

Evolutionary Ecology, Journal Year: 2025, Volume and Issue: unknown

Published: Jan. 15, 2025

Abstract Organisms require energy for survival, growth, and reproduction. In a system with finite supply, fluctuations in resource availability can select plasticity the allocation of resources between competing physiological processes. Infrequently-feeding snakes, which naturally experience extended episodes fasting, have evolved capacity to modulate gastrointestinal (GI) performance response changes digestive load. Specifically, gut be downregulated during fasting reduce metabolic maintenance costs. Some is, however, required upregulate again allow exploitation captured prey. Despite significant inquiry into relationship sit-and-wait foraging tactics GI plasticity, quantitative examination optimal strategy an infrequently-feeding snake is lacking. Here, we construct optimisation model quantitatively predict this terms length time post-feeding after downregulation occurs minimum prey mass initiate upregulation. Contrary long-held assertions, our simulations that snakes all sizes practice benefit from consuming relatively small We identify as adaptive when are encountered infrequently, assert it critical factor determining predator vulnerability food scarcity. When parameterising model, found distribution potential body poorly characterised literature. Accordingly, frequency individual within terrestrial community key focus future ecological research.

Language: Английский

---

Advancing Sika deer detection and distance estimation through comprehensive camera calibration and distortion analysis

Ecological Informatics, Journal Year: 2025, Volume and Issue: unknown, P. 103064 - 103064

Published: Feb. 1, 2025

Language: Английский

---

Fixed-Wing UAV Pose Estimation Using a Self-Organizing Map and Deep Learning

Robotics, Journal Year: 2024, Volume and Issue: 13(8), P. 114 - 114

Published: July 27, 2024

In many Unmanned Aerial Vehicle (UAV) operations, accurately estimating the UAV’s position and orientation over time is crucial for controlling its trajectory. This especially important when considering landing maneuver, where a ground-based camera system can estimate 3D orientation. A Red, Green, Blue (RGB) monocular approach be used this purpose, allowing more complex algorithms higher processing power. The proposed method uses hybrid Artificial Neural Network (ANN) model, incorporating Kohonen (KNN) or Self-Organizing Map (SOM) to identify feature points representing cluster obtained from binary image containing UAV. Deep (DNN) architecture then actual UAV pose based on single frame, including translation Utilizing Computer-Aided Design (CAD) network structure easily trained using synthetic dataset, fine-tuning done perform transfer learning deal with real data. experimental results demonstrate that achieves high accuracy, characterized by low errors in estimation. implementation paves way automating operational tasks like autonomous landing, which hazardous prone failure.

Language: Английский

---

Passenger Location Estimation in Public Transport: Evaluating Methods and Camera Placement Impact

IEEE Transactions on Intelligent Transportation Systems, Journal Year: 2024, Volume and Issue: 25(11), P. 17878 - 17887

Published: Aug. 6, 2024

Language: Английский

---

Distance Estimation Approach for Maritime Traffic Surveillance Using Instance Segmentation

Journal of Marine Science and Engineering, Journal Year: 2023, Volume and Issue: 12(1), P. 78 - 78

Published: Dec. 28, 2023

Maritime traffic monitoring systems are particularly important in Mediterranean ports, as they provide more comprehensive data collection compared to traditional such the Automatic Identification System (AIS), which is not mandatory for all vessels. This paper improves existing real-time maritime by introducing a distance estimation algorithm monocular cameras, aims high quality metadata density analysis. Two methods based on pinhole camera model presented: Vessel-Focused Distance Estimation (VFDE) and novel Vessel Object-Focused (VOFDE). While VFDE uses predefined height of vessel estimation, VOFDE standardized dimensions objects vessel, detected with Convolutional Neural Network (CNN) instance segmentation enhance accuracy. Our evaluation covers distances up 414 m, significantly beyond scope previous studies. When measured precise instrument, achieves Percentage Deviation Index (PDI) 1.34% 9.45%. advance holds significant potential improving surveillance cameras also applicable other areas, low-cost vehicles equipped single cameras.

Language: Английский

---