RFIDeep: Unfolding the potential of deep learning for radio‐frequency identification

Methods in Ecology and Evolution, Journal Year: 2023, Volume and Issue: 14(11), P. 2814 - 2826

Published: Aug. 22, 2023

Abstract Automatic monitoring of wildlife is becoming a critical tool in the field ecology. In particular, Radio‐Frequency IDentification (RFID) now widespread technology to assess phenology, breeding and survival many species. While RFID produces massive datasets, no established fast accurate methods are yet available for this type data processing. Deep learning approaches have been used overcome similar problems other scientific fields hence might hold potential these analytical challenges unlock full studies. We present deep workflow, coined “RFIDeep”, derive ecological features, such as status outcome, from mark‐recapture data. To demonstrate performance RFIDeep with complex we long‐term automatic long‐lived seabird that breeds densely packed colonies, daily entries exits. determine individual phenology each season, first developed one‐dimensional convolution neural network (1D‐CNN) architecture. Second, account variance technical limitations acquisition, built new augmentation step mimicking shift dates missing detections, common issue RFIDs. Third, identify segments activity during classification, also included visualisation tool, which allows users understand what usually considered “black box” learning. With three steps, achieved high accuracy all parameters: = 96.3%; phenological 86.9%; success 97.3%. has unfolded artificial intelligence tracking changes animal populations, multiplying benefit automated undisturbed populations. an open source code facilitate use, adaptation, or enhancement wide variety addition tremendous time saving analysing large our study shows capacities CNN models autonomously detect ecologically meaningful patterns through techniques, seldom

Language: Английский

---

Use of geolocators for investigating breeding ecology of a rock crevice‐nesting seabird: Method validation and impact assessment

Ecology and Evolution, Journal Year: 2023, Volume and Issue: 13(3)

Published: March 1, 2023

Abstract Investigating ecology of marine animals imposes a continuous challenge due to their temporal and/or spatial unavailability. Light‐based geolocators (GLS) are animal‐borne devices that provide relatively cheap and efficient method track seabird movement commonly used study migration. Here, we explore the potential GLS data establish individual behavior during breeding period in rock crevice‐nesting seabird, Little Auk, Alle alle . By deploying on 12 pairs, developed methodological workflow extract birds' from (nest attendance, colony foraging activity), validated its accuracy using extracted well‐established based video recordings. We also compared outcome, as well behavioral patterns logged individuals with control group treated similarly all aspects except for deployment logger, assess short‐term logger effects fitness behavior. found high GLS‐established patterns, especially incubation early chick rearing (when birds spend long time nest). observed no apparent effect outcome but recorded some changes (longer bouts shorter trips). Our provides useful framework establishing attendance foraging) (light conductivity), period. Given does not seem affect fine‐scale behavior, our is likely be applicable variety crevice/burrow nesting seabirds, even though precautions should taken reduce effect. Finally, because each species may have own ecological specificity, recommend performing pilot before implementing new system.

Language: Английский

---

Evaluating the accuracy and biological meaning of visits to RFID‐enabled bird feeders using video

Ecology and Evolution, Journal Year: 2021, Volume and Issue: 11(23), P. 17132 - 17141

Published: Nov. 16, 2021

Abstract Radio‐frequency identification (RFID) technology has gained popularity in ornithological studies as a way to collect large quantities of data answer specific biological questions, but few published report methodologies used for validating the accuracy RFID data. Further, connections between and behaviors interest study are not always clearly established. These methodological deficiencies may seriously impact study's results subsequent interpretation. We built RFID‐equipped bird feeders mounted them at three sites Tompkins County, New York. deployed passive integrated transponder tags on black‐capped chickadees, tufted titmice, white‐breasted nuthatches GoPro video camera record tagged species feeders. then reviewed determine reader understand birds’ behavior found that our system recorded only 34.2% all visits by birds ( n = 237) detection increased with length visit. also two other visited feeders, American goldfinch hairy woodpecker, retrieved food 79.5% their visits. Chickadees, nuthatches, woodpeckers spent, average, 2.3 s one seed per In contrast, goldfinches spent an average 9.0 consumed up 30 seeds Our demonstrate importance confirming can be identify behavioral characteristics associated reader's detections. This simple — yet time‐intensive method assessing meaning is useful research focusing various taxa systems.

Language: Английский

---

Animal-friendly behavioral testing in field studies: examples from ground squirrels

Frontiers in Behavioral Neuroscience, Journal Year: 2023, Volume and Issue: 17

Published: Aug. 23, 2023

Field studies of behavior provide insight into the expression in its natural ecological context and can serve as an important complement to behavioral conducted lab under controlled conditions. In addition naturalistic observations, testing be component field behavior. This mini review evaluates a sample methods identify ways which animal-friendly generate ethologically relevant data. Specific examples, primarily from ground squirrels, are presented illustrate principles applied guide methods. Tests with animals their habitat that elicit naturally occurring responses minimize stress disturbance for animals, well disruption larger ecosystem, have high ethological validity. When trapped or handled part study, incorporated handling procedures reduce overall disturbance. is evaluated arena, arena designed resemble conditions increase relevance test. Efforts time spent arenas also animals. Adapting test species facilitate reduced subjects increased

Language: Английский

---

RFIDeep: Unfolding the Potential of Deep Learning for Radio-Frequency Identification

bioRxiv (Cold Spring Harbor Laboratory), Journal Year: 2023, Volume and Issue: unknown

Published: March 27, 2023

Abstract Automatic monitoring of wildlife is becoming a critical tool in the field ecology. In particular, Radio-Frequency IDentification (RFID) now widespread technology to assess phenology, breeding, and survival many species. While RFID produces massive datasets, no established fast accurate methods are yet available for this type data processing. Deep learning approaches have been used overcome similar problems other scientific fields hence might hold potential these analytical challenges unlock full studies. We present deep workflow, coined “RFIDeep”, derive ecological features, such as breeding status outcome, from mark-recapture data. To demonstrate performance RFIDeep with complex we long-term automatic long-lived seabird that breeds densely packed colonies, daily entries exits. determine individual phenology each season, first developed one-dimensional convolution neural network (1D-CNN) architecture. Second, account variance technical limitations acquisition, built new augmentation step mimicking shift dates missing detections, common issue RFIDs. Third, identify segments activity during classification, also included visualisation tool, which allows users understand what usually considered “black box” learning. With three steps, achieved high accuracy all parameters: = 96.3%; phenological 86.9%; success 97.3%. has unfolded artificial intelligence tracking changes animal populations, multiplying benefit automated undisturbed populations. an open source code facilitate use, adaptation, or enhancement wide variety addition tremendous time saving analyzing large our study shows capacities CNN models autonomously detect ecologically meaningful patterns through techniques, seldom

Language: Английский

---

Automated tracking of avian parental care behavior

Published: Nov. 16, 2023

1. Parental care may be an important source of phenotypic variation for ecological and evolutionary processes. However, it can difficult to collect interpret data on parental behaviors. To address these challenges, we developed a new hardware software platform automated behavioral tracking called ABISSMAL (Automated Behavioral Tracking by Integrating Sensors that Survey Movements Around target Location).2. automatically collects across low-cost sensors with built-in system monitoring error logging. also generates inferences internal validation integrating multiple movement sensors.3. We successfully used track nest attendance activities performed captive zebra finches (Taeniopygia guttata) raised chicks through fledging. highlight the derive from integrated datasets represent discrete events, including types behaviors, direction magnitude movements.4. streamlines process collection, curation, interpretation researchers studying many experimental replicates over long developmental timescales. is modular deployed different combinations suit research questions setups. made open-access GitHub detailed documentation facilitate widespread use modification.

Language: Английский

---