Introducing a Central African Primate Vocalisation Dataset for Automated Species Classification

Zwerts, Joeri A.; Treep, Jelle; Kaandorp, Casper S.; Meewis, Floor; Koot, Amparo C.; Kaya, Heysem

doi:10.21437/interspeech.2021-154

Cited by 4 publications

(9 citation statements)

References 22 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…These classifiers are not yet available for most species in tropical forests. Exceptions include elephants (Wrege et al, 2017), some primates (Heinicke et al, 2015;Zwerts et al, 2021), and birds (Priyadarshani et al, 2018). They do exist for gunshots, which can be used for eco-surveillance purposes (Astaras et al, 2017).…”

Section: Passive Acoustic Monitoringmentioning

confidence: 99%

Methods for wildlife monitoring in tropical forests: Comparing human observations, camera traps, and passive acoustic sensors

Zwerts

Stephenson

Maisels

et al. 2021

Conservat Sci and Prac

Self Cite

View full text Add to dashboard Cite

Wildlife monitoring is essential for conservation science and data‐driven decision‐making. Tropical forests pose a particularly challenging environment for monitoring wildlife due to the dense vegetation, and diverse and cryptic species with relatively low abundances. The most commonly used monitoring methods in tropical forests are observations made by humans (visual or acoustic), camera traps, or passive acoustic sensors. These methods come with trade‐offs in terms of species coverage, accuracy and precision of population metrics, available technical expertise, and costs. Yet, there are no reviews that compare the characteristics of these methods in detail. Here, we comprehensively review the advantages and limitations of the three mentioned methods, by asking four key questions that are always important in relation to wildlife monitoring: (1) What are the target species?; (2) Which population metrics are desirable and attainable?; (3) What expertise, tools, and effort are required for species identification?; and (4) Which financial and human resources are required for data collection and processing? Given the diversity of monitoring objectives and circumstances, we do not aim to conclusively prescribe particular methods for all situations. Neither do we claim that any one method is superior to others. Rather, our review aims to support scientists and conservation practitioners in understanding the options and criteria that must be considered in choosing the appropriate method, given the objectives of their wildlife monitoring efforts and resources available. We focus on tropical forests because of their high conservation priority, although the information put forward is also relevant for other biomes.

show abstract

Section: Passive Acoustic Monitoringmentioning

confidence: 99%

Methods for wildlife monitoring in tropical forests: Comparing human observations, camera traps, and passive acoustic sensors

Zwerts

Stephenson

Maisels

et al. 2021

Conservat Sci and Prac

Self Cite

View full text Add to dashboard Cite

show abstract

“…Multi-class species classification was conducted between Blue-winged and Golden-winged warbler (see Table 2). In addition the Computational Paralinguistics Challenge Primate (ComParE-PRS) 37,38 dataset was used to distinguish between four different primate species (see Table 4). The dataset includes over 10,000 annotated vocalizations from Chimpanzees (Pan troglodytes), Mandrills (Mandrillus sphinx), Red-capped mangabeys (Cercocebus torquatus), and a mixed group of Guenons (Cercopithecus spp.).…”

Section: Bioacoustic Data Materialmentioning

confidence: 99%

“…The dataset includes over 10,000 annotated vocalizations from Chimpanzees (Pan troglodytes), Mandrills (Mandrillus sphinx), Red-capped mangabeys (Cercocebus torquatus), and a mixed group of Guenons (Cercopithecus spp.). Additionally, exactly the same number of noise samples as vocalizations were extracted to make up the noise class 38 . Multi-class call type classification was computed for three main call type classes of monk parakeet vocalizations including alarm, contact, and other calls, listed in Table 3.…”

Section: Bioacoustic Data Materialmentioning

confidence: 99%

“…Regarding the two challenge datasets-BirdVox-Full-Night 36 and ComParE-PRS 37,38 -the original predefined data splits have been applied for reasons of comparison. Network training and evaluation was computed utilizing mid-range graphics processing units (GPUs) (e.g., Nvidia GTX 1080), as well as standard central processing units (CPUs), showing the broad applicability of the training setup.…”

Section: Animal-spotmentioning

confidence: 99%

“…The first evaluation scenario visualizes the following training, validation, and test metrics: accuracy (ACC), true-positive-rate (TPR), false-positive-rate (FPR), precision (PREC), F1-score (F1), and area under the ROC curve (AUC). In case of the ComParE-PRS 37,38 primate species recognition challenge, only the unweighted average recall (UAR) was reported due to comparability reasons.…”

Section: Animal-spotmentioning

confidence: 99%

See 2 more Smart Citations

ANIMAL-SPOT enables animal-independent signal detection and classification using deep learning

Bergler

Smeele

Tyndel

et al. 2022

Sci Rep

View full text Add to dashboard Cite

Bioacoustic research spans a wide range of biological questions and applications, relying on identification of target species or smaller acoustic units, such as distinct call types. However, manually identifying the signal of interest is time-intensive, error-prone, and becomes unfeasible with large data volumes. Therefore, machine-driven algorithms are increasingly applied to various bioacoustic signal identification challenges. Nevertheless, biologists still have major difficulties trying to transfer existing animal- and/or scenario-related machine learning approaches to their specific animal datasets and scientific questions. This study presents an animal-independent, open-source deep learning framework, along with a detailed user guide. Three signal identification tasks, commonly encountered in bioacoustics research, were investigated: (1) target signal vs. background noise detection, (2) species classification, and (3) call type categorization. ANIMAL-SPOT successfully segmented human-annotated target signals in data volumes representing 10 distinct animal species and 1 additional genus, resulting in a mean test accuracy of 97.9%, together with an average area under the ROC curve (AUC) of 95.9%, when predicting on unseen recordings. Moreover, an average segmentation accuracy and F1-score of 95.4% was achieved on the publicly available BirdVox-Full-Night data corpus. In addition, multi-class species and call type classification resulted in 96.6% and 92.7% accuracy on unseen test data, as well as 95.2% and 88.4% regarding previous animal-specific machine-based detection excerpts. Furthermore, an Unweighted Average Recall (UAR) of 89.3% outperformed the multi-species classification baseline system of the ComParE 2021 Primate Sub-Challenge. Besides animal independence, ANIMAL-SPOT does not rely on expert knowledge or special computing resources, thereby making deep-learning-based bioacoustic signal identification accessible to a broad audience.

show abstract

An integrated passive acoustic monitoring and deep learning pipeline for black‐and‐white ruffed lemurs (Varecia variegata) in Ranomafana National Park, Madagascar

Batist,

Dufourq,

Jeantet

et al. 2024

American J Primatol

View full text Add to dashboard Cite

The urgent need for effective wildlife monitoring solutions in the face of global biodiversity loss has resulted in the emergence of conservation technologies such as passive acoustic monitoring (PAM). While PAM has been extensively used for marine mammals, birds, and bats, its application to primates is limited. Black‐and‐white ruffed lemurs (Varecia variegata) are a promising species to test PAM with due to their distinctive and loud roar‐shrieks. Furthermore, these lemurs are challenging to monitor via traditional methods due to their fragmented and often unpredictable distribution in Madagascar's dense eastern rainforests. Our goal in this study was to develop a machine learning pipeline for automated call detection from PAM data, compare the effectiveness of PAM versus in‐person observations, and investigate diel patterns in lemur vocal behavior. We did this study at Mangevo, Ranomafana National Park by concurrently conducting focal follows and deploying autonomous recorders in May–July 2019. We used transfer learning to build a convolutional neural network (optimized for recall) that automated the detection of lemur calls (57‐h runtime; recall = 0.94, F1 = 0.70). We found that PAM outperformed in‐person observations, saving time, money, and labor while also providing re‐analyzable data. Using PAM yielded novel insights into V. variegata diel vocal patterns; we present the first published evidence of nocturnal calling. We developed a graphic user interface and open‐sourced data and code, to serve as a resource for primatologists interested in implementing PAM and machine learning. By leveraging the potential of this pipeline, we can address the urgent need for effective primate population surveys to inform conservation strategies.

show abstract

Introducing a Central African Primate Vocalisation Dataset for Automated Species Classification

Cited by 4 publications

References 22 publications

Methods for wildlife monitoring in tropical forests: Comparing human observations, camera traps, and passive acoustic sensors

Methods for wildlife monitoring in tropical forests: Comparing human observations, camera traps, and passive acoustic sensors

ANIMAL-SPOT enables animal-independent signal detection and classification using deep learning

An integrated passive acoustic monitoring and deep learning pipeline for black‐and‐white ruffed lemurs (Varecia variegata) in Ranomafana National Park, Madagascar

Contact Info

Product

Resources

About