BirdNET: A deep learning solution for avian diversity monitoring

Ecological Informatics, Год журнала: 2021, Номер 61, С. 101236 - 101236

Опубликована: Янв. 27, 2021

Variation in avian diversity space and time is commonly used as a metric to assess environmental changes. Conventionally, such data were collected by expert observers, but passively acoustic rapidly emerging an alternative survey technique. However, efficiently extracting accurate species richness from large audio datasets has proven challenging. Recent advances deep artificial neural networks (DNNs) have transformed the field of machine learning, frequently outperforming traditional signal processing techniques domain event detection classification. We developed DNN, called BirdNET, capable identifying 984 North American European bird sound. Our task-specific model architecture was derived family residual (ResNets), consisted 157 layers with more than 27 million parameters, trained using extensive pre-processing, augmentation, mixup. tested against three independent datasets: (a) 22,960 single-species recordings; (b) 286 h fully annotated soundscape array autonomous recording units design analogous what researchers might use measure setting; (c) 33,670 single high-quality omnidirectional microphone deployed near four eBird hotspots frequented birders. found that domain-specific augmentation key build models are robust high ambient noise levels can cope overlapping vocalizations. Task-specific designs training regimes for recognition perform on-par very complex architectures other domains (e.g., object images). also temporal resolution input spectrograms (short FFT window length) improves classification performance sounds. In summary, BirdNET achieved mean average precision 0.791 recordings, F0.5 score 0.414 soundscapes, correlation 0.251 hotspot observation across 121 4 years data. By enabling efficient extraction vocalizations many hundreds potentially vast amounts data, similar tools potential add tremendous value existing future may transform ecology conservation.

Язык: Английский

---

Automatic acoustic detection of birds through deep learning: The first Bird Audio Detection challenge

Methods in Ecology and Evolution, Год журнала: 2018, Номер 10(3), С. 368 - 380

Опубликована: Окт. 10, 2018

Abstract Assessing the presence and abundance of birds is important for monitoring specific species as well overall ecosystem health. Many are most readily detected by their sounds, thus, passive acoustic highly appropriate. Yet often held back practical limitations such need manual configuration, reliance on example sound libraries, low accuracy, robustness, limited ability to generalise novel conditions. Here, we report outcomes from a collaborative data challenge. We present new datasets, summarise machine learning techniques proposed challenge teams, conduct detailed performance evaluation, discuss how approaches detection can be integrated into remote projects. Multiple methods were able attain around 88% area under receiver operating characteristic (ROC) curve (AUC), much higher than previous general‐purpose methods. With modern learning, including deep bird achieve very high retrieval rates in data, with no recalibration, pretraining detector target or conditions environment.

Язык: Английский

---

Automated birdsong recognition in complex acoustic environments: a review

Journal of Avian Biology, Год журнала: 2018, Номер 49(5)

Опубликована: Янв. 10, 2018

Conservationists are increasingly using autonomous acoustic recorders to determine the presence/absence and abundance of bird species. Unlike humans, these can be left in field for extensive periods time any habitat. Although data acquisition is automated, manual processing recordings labour intensive, tedious, prone bias due observer variations. Hence automated birdsong recognition an efficient alternative. However, only few ecologists conservationists utilise existing recognisers process unattended because software calibration exceptionally high requires considerable knowledge signal underlying systems, making tools less user‐friendly. Even allowing difficulties, getting accurate results exceedingly hard. In this review we examine state‐of‐the‐art, summarising discussing methods currently available each essential parts a recogniser, also software. The key reasons behind poor that very noisy, calls from birds long way recorder faint or corrupted, there overlapping many different birds. addition, large numbers species calling one recording, therefore method has scale species, at least avoid misclassifying another as particular interest. We found areas importance, particularly question noise reduction, amongst researched. cases where individual essential, such conservation work, suggest specialised (species‐specific) passive monitoring required. believe it important comparable measures, datasets, used enable compared.

Язык: Английский

---

Autonomous sound recording outperforms human observation for sampling birds: a systematic map and user guide

Ecological Applications, Год журнала: 2019, Номер 29(6)

Опубликована: Июнь 17, 2019

Abstract Autonomous sound recording techniques have gained considerable traction in the last decade, but question remains whether they can replace human observation surveys to sample sonant animals. For birds particular, survey methods been tested extensively using point counts and surveys. Here, we review latest evidence for this taxon within frame of a systematic map. We compare sampling effectiveness these two methods, output variables produce, their practicality. When assessed against standard counts, autonomous proves be powerful tool that samples at least as many species. This technology monitor an exhaustive, standardized, verifiable way. Moreover, recorders give access entire soundscapes from which new data types derived (vocal activity, acoustic indices). Variables such abundance, density, occupancy, or species richness obtained yield sets are comparable compatible with counts. Finally, allow investigations high temporal spatial resolution coverage, more cost effective cannot achieved by observations alone, even though small‐scale studies might when carried out Sound deployed places, scalable reliable, making them better choice bird increasingly data‐driven time. provide overview currently available discuss specifications guide future study designs.

Язык: Английский

---

Acoustic indices as proxies for biodiversity: a meta‐analysis

Biological reviews/Biological reviews of the Cambridge Philosophical Society, Год журнала: 2022, Номер 97(6), С. 2209 - 2236

Опубликована: Авг. 17, 2022

ABSTRACT As biodiversity decreases worldwide, the development of effective techniques to track changes in ecological communities becomes an urgent challenge. Together with other emerging methods ecology, acoustic indices are increasingly being used as novel tools for rapid assessment. These based on mathematical formulae that summarise features audio samples, aim extracting meaningful information from soundscapes. However, application this automated method has revealed conflicting results across literature, conceptual and empirical controversies regarding its primary assumption: a correlation between biological diversity. After more than decade research, we still lack statistically informed synthesis power elucidates whether they effectively function proxies Here, reviewed studies testing relationship diversity metrics (species abundance, species richness, diversity, abundance sounds, sounds) 11 most commonly indices. From 34 studies, extracted 364 effect sizes quantified magnitude direct link estimates conducted meta‐analysis. Overall, had moderate positive ( r = 0.33, CI [0.23, 0.43]), showed inconsistent performance, highly variable both within among studies. Over time, have been disregarding validation those examining progressively reporting smaller sizes. Some studied [acoustic entropy index (H), normalised difference soundscape (NDSI), complexity (ACI)] performed better retrieving information, sounds (number identified or unidentified species) best estimated facet local communities. We found no type monitored environment (terrestrial versus aquatic) procedure (acoustic non‐acoustic) performance indices, suggesting certain potential generalise their research contexts. also common statistical issues knowledge gaps remain be addressed future such high rate pseudoreplication multiple unexplored combinations metrics, taxa, regions. Our findings confirm limitations efficiently quantify alpha highlight caution is necessary when using them surrogates especially if employed single predictors. Although these able partially capture endorsing some extent rationale behind promising bases developments, far biodiversity. To guide efficient use review principal theoretical practical shortcomings, well prospects challenges Altogether, provide first comprehensive overview relation pave way standardised monitoring.

Язык: Английский

---

Estimating bird density using passive acoustic monitoring: a review of methods and suggestions for further research

Ibis, Год журнала: 2021, Номер 163(3), С. 765 - 783

Опубликована: Фев. 8, 2021

Passive acoustic monitoring is a non‐invasive tool for automated wildlife monitoring. This technique has several advantages and addresses many of the biases related to traditional field surveys. However, locating animal sounds using autonomous recording units (ARUs) can be technically challenging therefore ARUs have traditionally been little employed estimate density. Nonetheless, approaches proposed in recent years carry out acoustic‐based bird density estimations. We conducted literature review studies that used estimating densities or abundances order describe applications improve future programmes. detected growing interest use last 6 (2014–19), with total 31 articles assessing topic. The most common approach was relationship between number vocalizations per time abundance estimated (61%). In 26 (79%), estimates obtained by human surveyors agreed those ARUs. Some proven able reduce surveys, such as considering imperfect detection (spatially explicit capture–recapture, microphone arrays), applying paired sampling control different radius humans ARUs, including relative sound level measurements allow researchers distance recorder. did not include any covariates existing some recorder, which may hamper comparisons ARU Future should measurement recorder obtain estimations Finally, we provide guidelines applicability infer population studies.

Язык: Английский

---

BirdNET: applications, performance, pitfalls and future opportunities

Ibis, Год журнала: 2023, Номер 165(3), С. 1068 - 1075

Опубликована: Фев. 27, 2023

Automated recognition software is paramount for effective passive acoustic monitoring. BirdNET a free and recently developed bird sound recognizer. I performed literature review to evaluate the current applications performance of BirdNET, which growing in popularity but has been subject few assessments, provide recommendations future studies using BirdNET. Prior research employed wide range purposes have linked detections ecological processes or real‐world monitoring schemes. Among evaluated studies, average precision (% correctly identified) usually ranged around 72–85%, recall rate target species vocalizations detected) 33–84%. Some did not assess performance, hampers interpretation results may poorly informed decisions. Recommendations on how efficiency are provided. The impact confidence score threshold, user‐selected parameter as minimum reported, output although variable among consistent. use high thresholds increases percentage classified lowers proportion calls detected. selection an optimal depend priorities user goals. great tool automated it should be used with caution due inherent challenges identification. continued refinement suggests further improvements coming years.

Язык: Английский

---

Assessing the potential of BirdNET to infer European bird communities from large-scale ecoacoustic data

Ecological Indicators, Год журнала: 2024, Номер 164, С. 112146 - 112146

Опубликована: Май 20, 2024

Passive acoustic monitoring has become increasingly popular as a practical and cost-effective way of obtaining highly reliable data in ecological research projects. Increased ease collecting these means that, currently, the main bottleneck ecoacoustic projects is often time required for manual analysis passively collected recordings. In this study we evaluate potential current limitations BirdNET-Analyzer v2.4, most advanced generic deep learning algorithm bird recognition to date, tool assess community composition through automated large-scale data. To end, 3 datasets comprising total 629 environmental soundscapes 194 different sites spread across 19° latitude span Europe. We analyze using both BirdNET listening by local expert birders, then compare results obtained two methods performance at level each single vocalization entire recording sequences (1, 5 or 10 min). Since provides confidence score identification, minimum thresholds can be used filter out identifications with low scores, thus retaining only ones. The volume did not allow us estimate species-specific taxa, so instead evaluated global selected optimized when consistently applied all species. Our analyses reveal that if sufficiently high threshold used. However, inevitable trade-off between precision recall does obtain satisfactory metrics same time. found F1-scores remain moderate (<0.5) studied, extended duration seem currently necessary provide minimally comprehensive picture target community. estimate, however, usage species- context-specific would substantially improve benchmarks study. conclude judicious use AI-based provided represent powerful method assist assessment data, especially duration.

Язык: Английский

---

Comparing the sampling performance of sound recorders versus point counts in bird surveys: A meta‐analysis

Journal of Applied Ecology, Год журнала: 2018, Номер 55(6), С. 2575 - 2586

Опубликована: Июнь 29, 2018

Abstract Autonomous sound recording is a promising sampling method for birds and other vocalizing terrestrial wildlife. However, while there are clear advantages of passive acoustic monitoring methods over classical point counts conducted by humans, it has been difficult to quantitatively assess how they compare in their performance. Quantitative comparisons species richness between recorders human bird surveys have previously hampered the differing often unknown detection ranges or spaces among methods. We performed two meta‐analyses based on 28 studies where were paired with recordings at same sites. compared alpha gamma estimated both survey after equalizing effective ranges. further assessed influence technical specifications (microphone signal‐to‐noise ratio, height number) performance unlimited radius counts. show that standardizing ranges, from statistically indistinguishable, might be an avoidance effect Furthermore, we microphone ratio (a measure its quality), number positively affect through increasing range, allowing match Synthesis applications . demonstrate when used properly, high‐end systems can sample wildlife just as well observers conducting Correspondingly, suggest first standard methodology autonomous obtain results comparable enable practical sampling. also give recommendations carrying out making most recorders.

Язык: Английский

---

It's time to listen: there is much to be learned from the sounds of tropical ecosystems

Biotropica, Год журнала: 2018, Номер 50(5), С. 713 - 718

Опубликована: Июль 22, 2018

Abstract Knowledge that can be gained from acoustic data collection in tropical ecosystems is low‐hanging fruit. There every reason to record and with day, there are fewer excuses not do it. In recent years, the cost of recorders has decreased substantially (some purchased for under US $50, e.g., Hill et al . 2018) technology needed store analyze continuously improving (e.g., Corrada Bravo 2017, Xie 2017). Soundscape recordings provide a permanent site at given time contain wealth invaluable irreplaceable information. Although challenges remain, failure collect now would represent future generations researchers citizens benefit ecological research. this commentary, we (1) argue need increase monitoring systems; (2) describe types research questions conservation issues addressed passive ( PAM ) using both short‐ long‐term terrestrial freshwater habitats; (3) present an initial plan establishing global repository recordings.

Язык: Английский

---