Pith. sign in

REVIEW 2 major objections 5 minor 1 cited by

The iNaturalist Sounds Dataset

T0 review · 2 major / 5 minor · reviewed 2026-08-07 · deepseek-v4-flash

Challenge this review Re-run · record.json Download PDF Read on arXiv ↗

Pith's one-line read The iNaturalist Sounds Dataset assembles 230,000 weakly labeled animal recordings covering more than 5,500 species and shows that models trained on it transfer to strongly labeled bioacoustic benchmarks without fine-tuning.

desk verdict A genuinely useful large-scale weakly labeled animal audio dataset with honest benchmarking; the label-quality caveat is real but does not sink the core transfer claim. read the letter →

arxiv 2506.00343 v1 pith:36VF3WGI submitted 2025-05-31 cs.SD cs.LGeess.AS

Mustafa Chasmai , Alexander Shepard , Subhransu Maji , Grant Van Horn This is my paper
classification cs.SDcs.LGeess.AS
keywords iNatSoundsbioacousticsweaklylabeledlearninganimalsoundclassificationpretrainingtransfermultilabelcitizenscienceaudiogeographicpriors
verification ladder T0 review T1 audit T2 compute T3 formal

The pith

A machine-rendered reading of the paper's core claim, the machinery that carries it, and where it could break.

The reading

The paper builds iNatSounds, a collection of 230,000 audio recordings covering more than 5,500 animal species, all drawn from iNaturalist observations that carry a single weak, file-level species label. The authors' central claim is that even with such weak labels, models trained on iNatSounds learn useful bioacoustic representations: pretrained models transfer to strongly labeled bird and frog benchmarks without any fine-tuning, and adding a geographic prior raises iNatSounds top-1 accuracy into the low 60s. This matters because no comparably large, freely downloadable fine-grained animal audio dataset exists, and the resource could support conservation monitoring and public engagement tools. The paper also reports that multilabel objectives transfer better than multiclass objectives to strongly labeled datasets, while multiclass training scores higher on iNatSounds itself.

What carries the argument

The machinery is a weakly supervised audio pipeline. Audio is converted to mel-spectrogram images with a Short-Time Fourier Transform, and each recording's single positive label is treated as the label for randomly sampled 3-second windows. The paper compares multiclass cross-entropy ($\mathcal{L}_{\mathrm{CE}}$), an assume-negative multilabel loss ($\mathcal{L}_{\mathrm{AN}}$) derived from single-positive-label learning, and a mean-teacher loss ($\mathcal{L}_{\mathrm{ST}}$) that pseudo-labels augmented windows; evaluation averages predictions over strided windows and can mask species using SINR geographic range priors. The combination of weak labels, windowed training, and geo-filtering is what makes transfer possible without fine-tuning.

What would settle it

Take a random sample of, say, 200 test recordings from iNatSounds, have expert bioacousticians annotate every audible species, and compare their annotations with the dataset labels; if more than a small fraction of labels are wrong, or if expert-verified recordings are classified substantially better than unverified ones, the reported accuracy and transfer numbers are inflated by label noise.

Watch

Extended reading notes

Core claim

The central discovery is that weakly labeled citizen-science audio at species level is a viable pretraining resource. Each iNatSounds recording carries only a single positive label for the whole file, yet models trained on the dataset can be applied directly to strongly labeled downstream datasets such as SSW60, Powdermill, SWAMP, and AnuraSet without fine-tuning. On iNatSounds itself, the best multiclass models reach roughly 60.7% class-averaged top-1 accuracy on the test set when geographic priors filter out species that cannot occur at a recording's location, leaving ample room for improvement. The paper also finds that a multilabel objective generally transfers better to strongly labeled datasets than a multiclass objective, even though multiclass training maximizes accuracy on iNatSounds, and it cautions that iNatSounds performance alone does not reliably rank models for strongly labeled bioacoustic tasks.

Load-bearing premise

The whole benchmark rests on the assumption that iNaturalist's research-grade, majority-vote species identifications correctly name the species in each audio file, because the paper does not independently audit the audio labels.

Editorial extensions

If this is right

  • Scientists can download one archive and get roughly 230,000 recordings across 5,569 species, far more species than existing animal audio benchmarks, and use the released baselines as a reference point.
  • Pretrained iNatSounds models can be applied directly to strongly labeled avian and anuran datasets without fine-tuning, giving immediate tools for biodiversity monitoring.
  • For practical use, multilabel objectives plus mean-teacher training are the better recipe for strongly labeled downstream tasks, even though multiclass training wins on iNatSounds itself.
  • Geographic priors are a cheap and large accuracy boost (e.g., ResNet-50 top-1 from 52.6 to 60.7 on the test set), and should be used whenever location metadata is available.
  • Researchers should report both iNatSounds and strongly labeled results, because iNatSounds ranking alone does not reliably predict downstream performance.

Reading between the lines

Editorial extensions of the paper, not claims the author makes directly.

  • Editorial inference: If the label-quality premise holds, iNatSounds may play the role for audio that the iNat2021 image dataset played for vision, providing large-scale fine-grained pretraining whose embeddings generalize to narrower tasks, including long-tail species.
  • Editorial inference: The transfer results are only demonstrated for birds and frogs; a direct testable extension is whether iNatSounds pretraining also helps on strongly labeled mammal, insect, or reptile datasets.
  • Editorial inference: Because the paper trusts community labels without independent audio audit, the benchmark numbers should be read as upper bounds; an expert annotation audit of test recordings would show how much label noise remains.
  • Editorial inference: Since location metadata helped at test time, training with range-aware objectives or geo-conditioned heads is a natural next step the paper leaves implicit.
Share X Bluesky LinkedIn Reddit HN

Editorial analysis

A structured set of objections, weighed in public.

Desk editor's note, referee report, and a circularity audit.

Referee Report

2 major / 5 minor

Summary. The paper introduces iNatSounds, a weakly labeled audio dataset of roughly 230,000 recordings from over 5,500 species, built from research-grade iNaturalist observations. It describes the construction pipeline (year-based train/validation/test splits, species selection, spatio-temporal subsampling), benchmarks three backbone architectures under multiclass and multilabel objectives, optionally uses a geographic prior to filter predictions, and evaluates zero-shot transfer to strongly labeled downstream datasets (SSW60, Powdermill, SWAMP, AnuraSet). Additional experiments cover multimodal fusion on SSW60, the BEANS benchmark, a BirdCLEF 2024 comparison, and ablations. The central claims are that iNatSounds fills the gap of a large-scale, easily downloadable fine-grained animal audio dataset and that, despite weak labels, it serves as a useful pretraining resource.

Significance. If the claims hold, the dataset is a substantial contribution: existing large audio resources are either coarse-grained or not directly downloadable, and existing bioacoustic datasets are much smaller in species coverage. The paper's strengths include the scale and taxonomic breadth of the resource, the simple and reproducible construction and evaluation protocol, three-seed runs with standard deviations, honest discussion of the pitfalls of weak-label evaluation, and direct transfer experiments on several external strongly labeled datasets. The public release with code and licensing information is a concrete and valuable contribution. The principal caveat is that the ground-truth labels are inherited from iNaturalist research-grade status without an audio-specific accuracy audit, so the benchmark numbers should be interpreted as provisional until label quality is quantified.

major comments (2)
  1. [Section 3 (Initial filtering) and Section 5.1 (Table 1)] The test labels used in Table 1 and Fig. 3 are iNaturalist research-grade identifications, but the paper does not provide any audio-specific validation of these labels. The only supporting evidence cited is [57], which is not summarized and is not shown to apply to audio media; research-grade status can be based on accompanying images or metadata, and the only audio-level preprocessing described is sampling-rate filtering and resampling. Because a nontrivial mislabel rate would directly bias the reported Top-1/mAP numbers, especially for rare species with very few test files, please add a label-quality audit (e.g., expert verification of a stratified random sample of the test set, or agreement with an independent strongly labeled source) and report the estimated label accuracy, or otherwise bound the impact of label noise on the headline metrics.
  2. [Section 5.2 (Table 2) and abstract] The claim that iNatSounds is a useful pretraining resource is supported by absolute transfer numbers, but the main downstream tables do not include a control condition such as the same models without iNatSounds pretraining, or models pretrained on an existing large audio dataset (e.g., AudioSet or VGGSound) under the same evaluation protocol. The only such comparison appears in the appendix for BirdCLEF (Table A5). Please add a no-pretraining or alternative-pretraining baseline to at least one of the main downstream evaluations, or state the claim more narrowly as demonstrating that iNatSounds enables zero-shot transfer rather than outperforming existing pretraining resources.
minor comments (5)
  1. [Section 3] The sentence 'See [57] for an analysis of the quality of these identifications' should briefly summarize the accuracy numbers and clarify whether the cited experiment covers audio observations or only images; as written, the reader cannot assess the relevance of the citation.
  2. [Section 4 (Incorporating geographic priors)] The geo-prior threshold of 0.1 is selected on the validation set and then applied to the test set; please report the sensitivity of Table 1 to this threshold or describe the selection procedure in a way that makes the potential optimism explicit.
  3. [Section 5.2 (Multimodal classification)] The sentence 'We achieve state-of-the-art audio-only and multimodal performance on SSW60' is not backed by a comparison table or explicit prior state-of-the-art numbers; please provide the relevant references and values.
  4. [Table A2] There is a typo in the model name 'RresNet50' in Table A2; it should read 'ResNet50'.
  5. [Table 2] The 'Reported SOTA' row and the dagger/asterisk footnotes in Table 2 are difficult to parse in the submitted text; please format the table so that the comparison points and their caveats are unambiguous.

Circularity Check

1 steps flagged · score 4.0 of 10

Geo-prior evaluation is contaminated by training on the same export that supplies the test labels; the core pretraining-transfer claim rests on external benchmarks and is not circular.

  1. fitted input called prediction [Section 4 (Incorporating geographic priors); Table 1]
    "We use the SINR framework [14] to train a range estimation model for the 5,569 species in iNatSounds using observation data from the same iNaturalist database export used to build iNatSounds. However, instead of using observations with audio, we use the significantly larger set of observations with images. ... For each recording, we use the associated location to generate a binary mask by thresholding SINR probability of each species (a threshold of 0.1 performs best on the validation set)."

    The same February 1, 2024 export supplies both the 2023 audio test observations and the image observations used to train SINR; any test observation that also has an image contributes its species label and location to SINR's training data. The geo-prior mask is therefore trained on the very labels it is later used to help predict, so the reported geo-prior improvement (e.g., ResNet-50 52.6 to 60.7 in Table 1) is partly forced by reinserting test labels rather than by an independent geographic prior. The 0.1 threshold is additionally selected on the validation split. This is a partial circularity in an auxiliary evaluation result, not in the main pretraining-transfer claim.

full rationale

The central claim, that iNatSounds is a useful pretraining resource, is validated on external strongly labeled datasets (SSW60, Powdermill, SWAMP, AnuraSet, BEANS, BirdCLEF) without fine-tuning, and the paper compares iNatSounds pretraining against ImageNet and from-scratch baselines. That part is self-contained and not circular. The dataset labels are inherited from iNaturalist research-grade community identifications; this is an external ground-truth assumption and a correctness risk, not a circular derivation. The main circularity concern is the geo-prior evaluation in Section 4 and Table 1: the SINR range model is trained on image observations from the same iNaturalist export that also produced the audio test set, so the geographic mask can encode the species labels of the very observations being tested, and the threshold is tuned on validation. This inflates the geo-prior numbers but does not affect the core transfer claim. Self-citations to SINR, SSW60, and iNat2021 are contextual references to published work, not load-bearing uniqueness arguments. Overall, there is one partial circular step in an auxiliary evaluation, while the principal dataset and transfer results remain independently benchmarked, so a score of 4 is appropriate.

Assumptions & free parameters 3 free parameters · 3 assumptions · 0 invented entities

The paper's central result rests on the accuracy of citizen-science labels, on treating whole-file single labels as weak supervision, and on a geographic prior trained on image observations from the same platform. These are acknowledged assumptions, not hidden ones. There are no newly invented physical entities. The free parameters are standard hyperparameters and a curation threshold.

free parameters (3)
  • Geo-prior threshold = 0.1
    Chosen on the validation set (Sec. 4) for the SINR-based species mask; affects all Table 1 test numbers with geo-prior.
  • Mean teacher decay and consistency weight = alpha=0.999, lambda=1000
    Hyperparameters for the student-teacher loss (Eq. 3); chosen by manual experimentation, no sensitivity analysis shown.
  • Minimum species observations for val/test inclusion = 5
    Curation threshold that defines the 1,212 evaluation species; a different threshold would change the benchmark composition.
assumptions (3)
  • domain assumption iNaturalist research-grade identifications are accurate enough for training and evaluation.
    Section 3 builds the dataset from research-grade observations identified to species, relying on the platform's majority-vote system; the paper cites an accuracy experiment [57] rather than auditing labels itself.
  • domain assumption A recording's single positive label can be used as a weak label for the whole file.
    Section 4 and Eq. 2 assume all unlabelled species are negative, knowing this is false; this is the acknowledged weak-label assumption.
  • domain assumption The SINR range model trained on image observations transfers to audio observations.
    Section 4 uses SINR [14], trained on the larger image observation set from the same iNaturalist export, to mask species at test time; if image-based ranges are biased, the geo-prior results are optimistic.

how reviews work

0 comments
Cite this review

Pith. "Pith review of The iNaturalist Sounds Dataset." pith.science (2026). https://pith.science/paper/36VF3WGI

@misc{pith2026250600343,
  author       = {Pith},
  title        = {Pith review of: The iNaturalist Sounds Dataset},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/36VF3WGI}},
  note         = {Machine review of arXiv:2506.00343}
}
read the original abstract

We present the iNaturalist Sounds Dataset (iNatSounds), a collection of 230,000 audio files capturing sounds from over 5,500 species, contributed by more than 27,000 recordists worldwide. The dataset encompasses sounds from birds, mammals, insects, reptiles, and amphibians, with audio and species labels derived from observations submitted to iNaturalist, a global citizen science platform. Each recording in the dataset varies in length and includes a single species annotation. We benchmark multiple backbone architectures, comparing multiclass classification objectives with multilabel objectives. Despite weak labeling, we demonstrate that iNatSounds serves as a useful pretraining resource by benchmarking it on strongly labeled downstream evaluation datasets. The dataset is available as a single, freely accessible archive, promoting accessibility and research in this important domain. We envision models trained on this data powering next-generation public engagement applications, and assisting biologists, ecologists, and land use managers in processing large audio collections, thereby contributing to the understanding of species compositions in diverse soundscapes.

Figures

Figures reproduced from arXiv: 2506.00343 by the authors.

Figure 1
Figure 1. Overview. iNatSounds includes observations from around the world, capturing sounds of diverse species across major taxa. To the right, we show how iNatSounds compares to other general and animal specific sound datasets in terms of total audio duration and label diversity. Abstract We present the iNaturalist Sounds Dataset (iNatSounds), a collection of 230,000 audio files capturing sounds from over 5,500 species, con… view at source ↗
Figure 2
Figure 2. Dataset Statistics. Distribution of the number of recordings for each species in the train, validation, and test splits of iNatSounds. Some frequent and less frequent species are also shown. Note that the training set has a larger number of species than the validation and test splits. train split. With this splitting strategy, 147K, 91K and 106K observations are eligible for the train, validation and test splits res… view at source ↗
Figure 3
Figure 3. Analysis of iNatSounds Performance. Left: Sampled top confusion pairs highlighting fine-grained acoustic challenges. Right: Class-averaged Top-1 accuracy on iNatSounds val set, binned by training set size. We show plots with and without geo-priors (used to filter out geographically irrelevant species, see Sec. 4) and explore outliers for classes with a high number of training recordings (200-500). Both figures are f… view at source ↗
Figures from the paper (2 more)
Figure 4
Figure 4. Figure 4: Multiclass vs Multilabel Score Visualisations. We compare the top 10 outputs from multiclass and multilabel MobileNet classifiers given 3-second inputs from weakly labeled (left) and strongly labeled (right) datasets. Box and bar colors are shared for the the strongly …
Figure 5
Figure 5. Figure 5: SSW60 Multimodal Fusion. Left: Top-1 accuracy (not class averaged) for different modalities from the SSW60 test dataset. Right: Top-1 accuracy (not class averaged) vs modality bitrate for video-only models and audio-only models. Both: DeIT for video, ViT multiclass for…

Discussion (0). Sign in to comment.

Forward citations

Cited by 1 Pith paper

Reviewed papers in the Pith corpus that reference this work. Sorted by Pith novelty score. Full citation record

  1. Deformation Driven Suction Cups: A Mechanics-Based Approach to Wearable Electronics

    physics.med-ph 2025-08 unverdicted novelty 5.0 of 10

    Suction adhesion on soft skin depends on cup geometry relative to substrate compliance: wide flat cups lose suction on skin, narrow tall domes retain volume and stick better.

Reference graph

Works this paper leans on

110 extracted references · 64 canonical work pages · cited by 1 Pith paper

  1. [57]

    iNaturalist Observation Accuracy Experiment (0.3), 2024

    Scott Loarie. iNaturalist Observation Accuracy Experiment (0.3), 2024. URL https://www. inaturalist.org/observation_accuracy_experiments/4

    work page 2024

  2. [1]

    Acoustic scene classification using parallel combination of lstm and cnn

    Soo Hyun Bae, In Kyu Choi, and Nam Soo Kim. Acoustic scene classification using parallel combination of lstm and cnn. In DCASE, pages 11–15, 2016

    2016

  3. [2]

    Whombat: An open-source annotation tool for machine learning development in bioacoustics

    Santiago Martinez Balvanera, Oisin Mac Aodha, Matthew J Weldy, Holly Pringle, Ella Browning, and Kate E Jones. Whombat: An open-source annotation tool for machine learning development in bioacoustics. arXiv preprint arXiv:2308.12688, 2023

    work page Pith review arXiv 2023

  4. [3]

    Improving geographically extensive acoustic survey designs for modeling species occurrence with imperfect detection and misidentification

    Katharine M Banner, Kathryn M Irvine, Thomas J Rodhouse, Wilson J Wright, Rogelio M Rodriguez, and Andrea R Litt. Improving geographically extensive acoustic survey designs for modeling species occurrence with imperfect detection and misidentification. Ecology and Evolution, 8(12):6144–6156, 2018

    2018

  5. [4]

    Birdsnap: Large-scale fine-grained visual categorization of birds

    Thomas Berg, Jiongxin Liu, Seung Woo Lee, Michelle L Alexander, David W Jacobs, and Peter N Belhumeur. Birdsnap: Large-scale fine-grained visual categorization of birds. In Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition, pages 2011–2018, 2014

    2011

  6. [5]

    Visual recognition with humans in the loop

    Steve Branson, Catherine Wah, Florian Schroff, Boris Babenko, Peter Welinder, Pietro Perona, and Serge Belongie. Visual recognition with humans in the loop. In European Conference on Computer Vision, pages 438–451. Springer, 2010

    2010

  7. [6]

    Bird species categorization using pose normalized deep convolutional nets

    Steve Branson, Grant Van Horn, Serge Belongie, and Pietro Perona. Bird species categorization using pose normalized deep convolutional nets. In British Machine Vision Conference (BMVC), 2014

    2014

  8. [7]

    Multi-label learning with incomplete class assignments

    Serhat Selcuk Bucak, Rong Jin, and Anil K Jain. Multi-label learning with incomplete class assignments. In Computer Vision and Pattern Recognition (CVPR), pages 2801–2808, 2011

    2011

Show all 110 references
  1. [8]

    A dataset for benchmarking neotropical anuran calls identification in passive acoustic monitoring

    Juan Sebastián Cañas, María Paula Toro-Gómez, Larissa Sayuri Moreira Sugai, Hernán Darío Benítez Re- strepo, Jorge Rudas, Breyner Posso Bautista, Luís Felipe Toledo, Simone Dena, Adão Henrique Rosa Domingos, Franco Leandro de Souza, et al. A dataset for benchmarking neotropica...

    2023

  2. [9]

    VGGSound: A large-scale audio- visual dataset

    Honglie Chen, Weidi Xie, Andrea Vedaldi, and Andrew Zisserman. VGGSound: A large-scale audio- visual dataset. In IEEE International Conference on Acoustics, Speech and Signal Processing (ICASSP), pages 721–725, 2020. 10

    2020

  3. [10]

    Convolutional recurrent neural networks for music classification

    Keunwoo Choi, György Fazekas, Mark Sandler, and Kyunghyun Cho. Convolutional recurrent neural networks for music classification. In IEEE International Conference on Acoustics, Speech and Signal processing (ICASSP), pages 2392–2396, 2017

    2017

  4. [11]

    An annotated set of audio recordings of eastern north american birds containing frequency, time, and species information, 2021

    Lauren M Chronister, Tessa A Rhinehart, Aidan Place, and Justin Kitzes. An annotated set of audio recordings of eastern north american birds containing frequency, time, and species information, 2021

    2021

  5. [12]

    A collection of fully-annotated soundscape recordings from the southern sierra nevada mountain range

    Mary Clapp, Stefan Kahl, Erik Meyer, Megan McKenna, Holger Klinck, and Gail Patricelli. A collection of fully-annotated soundscape recordings from the southern sierra nevada mountain range. URL https: // doi. org/ 10. 5281/ zenodo. 7525805, 2023

  6. [13]

    Multi-label learning from single positive labels

    Elijah Cole, Oisin Mac Aodha, Titouan Lorieul, Pietro Perona, Dan Morris, and Nebojsa Jojic. Multi-label learning from single positive labels. In Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition, pages 933–942, 2021

    2021

  7. [14]

    Spatial implicit neural representations for global-scale species mapping

    Elijah Cole, Grant Van Horn, Christian Lange, Alexander Shepard, Patrick Leary, Pietro Perona, Scott Loarie, and Oisin Mac Aodha. Spatial implicit neural representations for global-scale species mapping. In International Conference on Machine Learning, pages 6320–6342. PMLR, 2023

    2023

  8. [15]

    Chirping up the right tree: Incorporating biological taxonomies into deep bioacoustic classifiers

    Aurora Linh Cramer, Vincent Lostanlen, Andrew Farnsworth, Justin Salamon, and Juan Pablo Bello. Chirping up the right tree: Incorporating biological taxonomies into deep bioacoustic classifiers. In IEEE International Conference on Acoustics, Speech and Signal Processing (ICASS...

    2020

  9. [16]

    https://dcase.community/, 2024

    DCASE. https://dcase.community/, 2024. Accessed on Jun 1, 2024

    2024

  10. [17]

    Fma: A dataset for music analysis

    Michaël Defferrard, Kirell Benzi, Pierre Vandergheynst, and Xavier Bresson. Fma: A dataset for music analysis. arXiv preprint arXiv:1612.01840, 2016

    2016 arXiv

  11. [18]

    Imagenet: A large-scale hierarchical image database

    Jia Deng, Wei Dong, Richard Socher, Li-Jia Li, Kai Li, and Li Fei-Fei. Imagenet: A large-scale hierarchical image database. In 2009 IEEE conference on computer vision and pattern recognition, pages 248–255. Ieee, 2009

    2009

  12. [19]

    Audio retrieval with wavtext5k and clap training

    Soham Deshmukh, Benjamin Elizalde, and Huaming Wang. Audio retrieval with wavtext5k and clap training. arXiv preprint arXiv:2209.14275, 2022

    2022 arXiv

  13. [20]

    An image is worth 16x16 words: Transformers for image recognition at scale

    Alexey Dosovitskiy, Lucas Beyer, Alexander Kolesnikov, Dirk Weissenborn, Xiaohua Zhai, Thomas Unterthiner, Mostafa Dehghani, Matthias Minderer, Georg Heigold, Sylvain Gelly, et al. An image is worth 16x16 words: Transformers for image recognition at scale. In International Con...

    2020

  14. [21]

    Clotho: An audio captioning dataset

    Konstantinos Drossos, Samuel Lipping, and Tuomas Virtanen. Clotho: An audio captioning dataset. In IEEE International Conference on Acoustics, Speech and Signal Processing (ICASSP), pages 736–740, 2020

    2020

  15. [22]

    Clap learning audio concepts from natural language supervision

    Benjamin Elizalde, Soham Deshmukh, Mahmoud Al Ismail, and Huaming Wang. Clap learning audio concepts from natural language supervision. In IEEE International Conference on Acoustics, Speech and Signal Processing (ICASSP), pages 1–5, 2023

    2023

  16. [23]

    Neural audio synthesis of musical notes with wavenet autoencoders

    Jesse Engel, Cinjon Resnick, Adam Roberts, Sander Dieleman, Mohammad Norouzi, Douglas Eck, and Karen Simonyan. Neural audio synthesis of musical notes with wavenet autoencoders. In International Conference on Machine Learning, pages 1068–1077. PMLR, 2017

    2017

  17. [24]

    Fsd50k: an open dataset of human-labeled sound events

    Eduardo Fonseca, Xavier Favory, Jordi Pons, Frederic Font, and Xavier Serra. Fsd50k: an open dataset of human-labeled sound events. IEEE/ACM Transactions on Audio, Speech, and Language Processing, 30: 829–852, 2021

    2021

  18. [25]

    https://freesound.org/, 2024

    Freesound. https://freesound.org/, 2024. Accessed on Jun 1, 2024

    2024

  19. [26]

    https://labs.freesound.org/datasets/, 2024

    Freesound Labs Datasets. https://labs.freesound.org/datasets/, 2024. Accessed on Jun 1, 2024

    2024

  20. [27]

    Gemmeke, Daniel P

    Jort F. Gemmeke, Daniel P. W. Ellis, Dylan Freedman, Aren Jansen, Wade Lawrence, R. Channing Moore, Manoj Plakal, and Marvin Ritter. Audio set: An ontology and human-labeled dataset for audio events. In Proc. IEEE ICASSP 2017, New Orleans, LA, 2017

    2017

  21. [28]

    Global birdsong embeddings enable superior transfer learning for bioacoustic classification

    Burooj Ghani, Tom Denton, Stefan Kahl, and Holger Klinck. Global birdsong embeddings enable superior transfer learning for bioacoustic classification. Scientific Reports, 13(1):22876, 2023. 11

    2023

  22. [29]

    Aves: Animal vocalization encoder based on self-supervision

    Masato Hagiwara. Aves: Animal vocalization encoder based on self-supervision. In IEEE International Conference on Acoustics, Speech and Signal Processing (ICASSP), pages 1–5, 2023

    2023

  23. [30]

    Beans: The benchmark of animal sounds

    Masato Hagiwara, Benjamin Hoffman, Jen-Yu Liu, Maddie Cusimano, Felix Effenberger, and Katie Zacarian. Beans: The benchmark of animal sounds. In IEEE International Conference on Acoustics, Speech and Signal Processing (ICASSP), pages 1–5, 2023

    2023

  24. [31]

    Deep residual learning for image recognition

    Kaiming He, Xiangyu Zhang, Shaoqing Ren, and Jian Sun. Deep residual learning for image recognition. In Proceedings of the IEEE conference on computer vision and pattern recognition, pages 770–778, 2016

    2016

  25. [32]

    A new benchmark and approach for fine-grained cross-media retrieval

    Xiangteng He, Yuxin Peng, and Liu Xie. A new benchmark and approach for fine-grained cross-media retrieval. In Proceedings of the 27th ACM international conference on multimedia, pages 1740–1748, 2019

    2019

  26. [33]

    Alexander Hopping, Stefan Kahl, and Holger Klinck

    W. Alexander Hopping, Stefan Kahl, and Holger Klinck. A collection of fully-annotated sound- scape recordings from the southwestern amazon basin. URL https: // doi. org/ 10. 5281/ zenodo. 7079124 , 2022

    2022

  27. [34]

    Meta-learning in neural networks: A survey

    Timothy Hospedales, Antreas Antoniou, Paul Micaelli, and Amos Storkey. Meta-learning in neural networks: A survey. IEEE transactions on pattern analysis and machine intelligence, 44(9):5149–5169, 2021

    2021

  28. [35]

    Searching for mobilenetv3

    Andrew Howard, Mark Sandler, Grace Chu, Liang-Chieh Chen, Bo Chen, Mingxing Tan, Weijun Wang, Yukun Zhu, Ruoming Pang, Vijay Vasudevan, et al. Searching for mobilenetv3. InProceedings of the IEEE/CVF international conference on computer vision, pages 1314–1324, 2019

    2019

  29. [36]

    https://www.inaturalist.org, 2024

    iNaturalist. https://www.inaturalist.org, 2024. Accessed on Jun 1, 2024

    2024

  30. [37]

    Spatial transformer networks

    Max Jaderberg, Karen Simonyan, Andrew Zisserman, et al. Spatial transformer networks. Advances in neural information processing systems, 28, 2015

    2015

  31. [38]

    Cochlscene: Acquisition of acoustic scene data using crowdsourcing

    Il-Young Jeong and Jeongsoo Park. Cochlscene: Acquisition of acoustic scene data using crowdsourcing. In 2022 Asia-Pacific Signal and Information Processing Association Annual Summit and Conference (APSIPA ASC), pages 17–21. IEEE, 2022

    2022

  32. [39]

    Learning visual features from large weakly supervised data

    Armand Joulin, Laurens Van Der Maaten, Allan Jabri, and Nicolas Vasilache. Learning visual features from large weakly supervised data. In European Conference on Computer Vision (ECCV), pages 67–84, 2016

    2016

  33. [40]

    Birdnet: A deep learning solution for avian diversity monitoring

    Stefan Kahl, Connor M Wood, Maximilian Eibl, and Holger Klinck. Birdnet: A deep learning solution for avian diversity monitoring. Ecological Informatics, 61:101236, 2021

    2021

  34. [41]

    A collection of fully-annotated soundscape recordings from the northeastern united states, august 2022a

    Stefan Kahl, Russell Charif, and Holger Klinck. A collection of fully-annotated soundscape recordings from the northeastern united states, august 2022a. URL https: // doi. org/ 10. 5281/ zenodo. 7079380 , 2022

    2022

  35. [42]

    Wood, Philip Chaon, M

    Stefan Kahl, Connor M. Wood, Philip Chaon, M. Zachariah Peery, and Holger Klinck. A collection of fully-annotated soundscape recordings from the western united states. URL https: // doi. org/ 10. 5281/ zenodo. 7050014 , 2022

    2022

  36. [43]

    Slow-fast auditory streams for audio recognition

    Evangelos Kazakos, Arsha Nagrani, Andrew Zisserman, and Dima Damen. Slow-fast auditory streams for audio recognition. In IEEE International Conference on Acoustics, Speech and Signal Processing (ICASSP), pages 855–859, 2021

    2021

  37. [44]

    Audiocaps: Generating captions for audios in the wild

    Chris Dongjoo Kim, Byeongchang Kim, Hyunmin Lee, and Gunhee Kim. Audiocaps: Generating captions for audios in the wild. In Conference of the North American Chapter of the Association for Computational Linguistics: Human Language Technologies, Volume 1 (Long and Short Papers), ...

    2019

  38. [45]

    Expanding neon biodiversity surveys with new instrumentation and machine learning approaches

    Justin Kitzes, Rachael Blake, Sara Bombaci, Melissa Chapman, Sandra M Duran, Tao Huang, Maxwell B Joseph, Samuel Lapp, Sergio Marconi, William K Oestreich, et al. Expanding neon biodiversity surveys with new instrumentation and machine learning approaches. Ecosphere, 12(11):e0...

    2021

  39. [46]

    Birdclef 2024, 2024

    Holger Klinck, Maggie, Sohier Dane, Stefan Kahl, Tom Denton, and Vijay Ramesh. Birdclef 2024, 2024. URL https://kaggle.com/competitions/birdclef-2024

    2024

  40. [47]

    Audio retrieval with natural language queries: A benchmark study

    A Sophia Koepke, Andreea-Maria Oncescu, João F Henriques, Zeynep Akata, and Samuel Albanie. Audio retrieval with natural language queries: A benchmark study. IEEE Transactions on Multimedia, 25: 2675–2685, 2022. 12

    2022

  41. [48]

    Whittlesearch: Image search with relative attribute feedback

    Adriana Kovashka, Devi Parikh, and Kristen Grauman. Whittlesearch: Image search with relative attribute feedback. In IEEE Conference on Computer Vision and Pattern Recognition, pages 2973–2980, 2012

    2012

  42. [49]

    The unreasonable effectiveness of noisy data for fine-grained recognition

    Jonathan Krause, Benjamin Sapp, Andrew Howard, Howard Zhou, Alexander Toshev, Tom Duerig, James Philbin, and Li Fei-Fei. The unreasonable effectiveness of noisy data for fine-grained recognition. In European Conference on Computer Vision (ECCV), pages 301–320, 2016

    2016

  43. [50]

    Exploiting weakly supervised visual patterns to learn from partial annotations

    Kaustav Kundu and Joseph Tighe. Exploiting weakly supervised visual patterns to learn from partial annotations. Advances in Neural Information Processing Systems, 33:561–572, 2020

    2020

  44. [51]

    Active learning-based species range estimation

    Christian Lange, Elijah Cole, Grant Horn, and Oisin Mac Aodha. Active learning-based species range estimation. Advances in Neural Information Processing Systems, 36, 2024

    2024

  45. [52]

    Meta-learning with differen- tiable convex optimization

    Kwonjoon Lee, Subhransu Maji, Avinash Ravichandran, and Stefano Soatto. Meta-learning with differen- tiable convex optimization. In IEEE/CVF conference on computer vision and pattern recognition, pages 10657–10665, 2019

    2019

  46. [53]

    Integrating new technologies to broaden the scope of northern spotted owl monitoring and linkage with usda forest inventory data

    Damon B Lesmeister and Julianna MA Jenkins. Integrating new technologies to broaden the scope of northern spotted owl monitoring and linkage with usda forest inventory data. Frontiers in Forests and Global Change, 5:966978, 2022

    2022

  47. [54]

    https://www.imageclef.org/LifeCLEF, 2024

    LifeCLEF. https://www.imageclef.org/LifeCLEF, 2024. Accessed on Jun 1, 2024

    2024

  48. [55]

    Bilinear cnn models for fine-grained visual recognition

    Tsung-Yu Lin, Aruni RoyChowdhury, and Subhransu Maji. Bilinear cnn models for fine-grained visual recognition. In IEEE international conference on computer vision, pages 1449–1457, 2015

    2015

  49. [56]

    Clotho-aqa: A crowdsourced dataset for audio question answering

    Samuel Lipping, Parthasaarathy Sudarsanam, Konstantinos Drossos, and Tuomas Virtanen. Clotho-aqa: A crowdsourced dataset for audio question answering. In IEEE European Signal Processing Conference (EUSIPCO), pages 1140–1144, 2022

    2022

  50. [58]

    Birdvox- full-night: A dataset and benchmark for avian flight call detection

    Vincent Lostanlen, Justin Salamon, Andrew Farnsworth, Steve Kelling, and Juan Pablo Bello. Birdvox- full-night: A dataset and benchmark for avian flight call detection. In2018 IEEE International Conference on Acoustics, Speech and Signal Processing (ICASSP), pages 266–270. IEEE, 2018

    2018

  51. [59]

    Building naturalistic emotionally balanced speech corpus by retrieving emotional speech from existing podcast recordings

    Reza Lotfian and Carlos Busso. Building naturalistic emotionally balanced speech corpus by retrieving emotional speech from existing podcast recordings. IEEE Transactions on Affective Computing, 10(4): 471–483, 2017

    2017

  52. [60]

    Teaching categories to human learners with visual explanations

    Oisin Mac Aodha, Shihan Su, Yuxin Chen, Pietro Perona, and Yisong Yue. Teaching categories to human learners with visual explanations. In Conference on Computer Vision and Pattern Recognition (CVPR), 2018

    2018

  53. [61]

    Presence-only geographical priors for fine-grained image classification

    Oisin Mac Aodha, Elijah Cole, and Pietro Perona. Presence-only geographical priors for fine-grained image classification. In International Conference on Computer Vision (ICCV), 2019

    2019

  54. [62]

    https://www.macaulaylibrary.org, 2024

    Macaulay Library. https://www.macaulaylibrary.org, 2024. Accessed on Jun 1, 2024

    2024

  55. [63]

    Exploring the limits of weakly supervised pretraining

    Dhruv Mahajan, Ross Girshick, Vignesh Ramanathan, Kaiming He, Manohar Paluri, Yixuan Li, Ashwin Bharambe, and Laurens Van Der Maaten. Exploring the limits of weakly supervised pretraining. In European conference on computer vision (ECCV), pages 181–196, 2018

    2018

  56. [64]

    Fine-grained visual classification of aircraft

    Subhransu Maji, Esa Rahtu, Juho Kannala, Matthew Blaschko, and Andrea Vedaldi. Fine-grained visual classification of aircraft. arXiv preprint arXiv:1306.5151, 2013

    2013 arXiv

  57. [65]

    What is the ground truth? reliability of multi-annotator data for audio tagging

    Irene Martín-Morató and Annamaria Mesaros. What is the ground truth? reliability of multi-annotator data for audio tagging. In 2021 29th European Signal Processing Conference (EUSIPCO), pages 76–80. IEEE, 2021

    2021

  58. [66]

    Wavcaps: A chatgpt-assisted weakly-labelled audio captioning dataset for audio-language multimodal research

    Xinhao Mei, Chutong Meng, Haohe Liu, Qiuqiang Kong, Tom Ko, Chengqi Zhao, Mark D Plumbley, Yuexian Zou, and Wenwu Wang. Wavcaps: A chatgpt-assisted weakly-labelled audio captioning dataset for audio-language multimodal research. arXiv preprint arXiv:2303.17395, 2023

    2023 arXiv

  59. [67]

    https://merlin.allaboutbirds.org/sound-id/, 2024

    Merlin Sound ID. https://merlin.allaboutbirds.org/sound-id/, 2024. Accessed on Jun 1, 2024. 13

    2024

  60. [68]

    Mesaros, T

    A. Mesaros, T. Heittola, A. Diment, B. Elizalde, A. Shah, E. Vincent, B. Raj, and T. Virtanen. DCASE 2017 challenge setup: Tasks, datasets and baseline system. In Proceedings of the Detection and Classifi- cation of Acoustic Scenes and Events 2017 Workshop (DCASE2017), pages 8...

    2017

  61. [69]

    Nips4bplus: a richly annotated birdsong audio dataset

    Veronica Morfi, Yves Bas, Hanna Pamuła, Hervé Glotin, and Dan Stowell. Nips4bplus: a richly annotated birdsong audio dataset. PeerJ Computer Science, 5:e223, 2019

    2019

  62. [70]

    A collection of fully-annotated soundscape recordings from the island of hawaii

    Amanda Navine, Stefan Kahl, Ann Tanimoto-Johnson, Holger Klinck, and Patrick Hart. A collection of fully-annotated soundscape recordings from the island of hawaii. URL https: // doi. org/ 10. 5281/ zenodo. 7078499 , 2022

    2022

  63. [71]

    A visual vocabulary for flower classification

    M-E Nilsback and Andrew Zisserman. A visual vocabulary for flower classification. In Conference on Computer Vision and Pattern recognition (CVPR), volume 2, pages 1447–1454, 2006

    2006

  64. [72]

    Recurrent neural networks for polyphonic sound event detection in real life recordings

    Giambattista Parascandolo, Heikki Huttunen, and Tuomas Virtanen. Recurrent neural networks for polyphonic sound event detection in real life recordings. In IEEE international conference on acoustics, speech and signal processing (ICASSP), pages 6440–6444, 2016

    2016

  65. [73]

    Specaugment: A simple data augmentation method for automatic speech recognition

    Daniel S Park, William Chan, Yu Zhang, Chung-Cheng Chiu, Barret Zoph, Ekin D Cubuk, and Quoc V Le. Specaugment: A simple data augmentation method for automatic speech recognition. Interspeech 2019, 2019

    2019

  66. [74]

    Study of positional encoding approaches for audio spectrogram transformers

    Leonardo Pepino, Pablo Riera, and Luciana Ferrer. Study of positional encoding approaches for audio spectrogram transformers. In IEEE International Conference on Acoustics, Speech and Signal Processing (ICASSP), pages 3713–3717, 2022

    2022

  67. [75]

    Esc: Dataset for environmental sound classification

    Karol J Piczak. Esc: Dataset for environmental sound classification. In Proceedings of the 23rd ACM international conference on Multimedia, pages 1015–1018, 2015

    2015

  68. [76]

    Soundscape ecology: the science of sound in the landscape

    Bryan C Pijanowski, Luis J Villanueva-Rivera, Sarah L Dumyahn, Almo Farina, Bernie L Krause, Brian M Napoletano, Stuart H Gage, and Nadia Pieretti. Soundscape ecology: the science of sound in the landscape. BioScience, 61(3):203–216, 2011

    2011

  69. [77]

    Meld: A multimodal multi-party dataset for emotion recognition in conversations

    Soujanya Poria, Devamanyu Hazarika, Navonil Majumder, Gautam Naik, Erik Cambria, and Rada Mihalcea. Meld: A multimodal multi-party dataset for emotion recognition in conversations. arXiv preprint arXiv:1810.02508, 2018

    2018 arXiv

  70. [78]

    Birdset: A multi-task benchmark for classification in avian bioacoustics

    Lukas Rauch, Raphael Schwinger, Moritz Wirth, René Heinrich, Jonas Lange, Stefan Kahl, Bernhard Sick, Sven Tomforde, and Christoph Scholz. Birdset: A multi-task benchmark for classification in avian bioacoustics. arXiv preprint arXiv:2403.10380, 2024

    2024 arXiv

  71. [79]

    An automatic classifier of bat sonotypes around the world

    Charlotte Roemer, Jean-François Julien, Pélé Patrice Ahoudji, Jean-Marie Chassot, Mauricio Genta, Raphaël Colombo, German Botto, Carlos A Negreira, Bruno Agossou Djossa, Ros Kiri Ing, et al. An automatic classifier of bat sonotypes around the world. Methods in Ecology and Evol...

    2021

  72. [80]

    A dataset and taxonomy for urban sound research

    Justin Salamon, Christopher Jacoby, and Juan Pablo Bello. A dataset and taxonomy for urban sound research. In Proceedings of the 22nd ACM international conference on Multimedia, pages 1041–1044, 2014

    2014

  73. [81]

    Towards the automatic classification of avian flight calls for bioacoustic monitoring

    Justin Salamon, Juan Pablo Bello, Andrew Farnsworth, Matt Robbins, Sara Keen, Holger Klinck, and Steve Kelling. Towards the automatic classification of avian flight calls for bioacoustic monitoring. PloS one, 11(11):e0166866, 2016

    2016

  74. [82]

    Prototypical networks for few-shot learning

    Jake Snell, Kevin Swersky, and Richard Zemel. Prototypical networks for few-shot learning. Advances in neural information processing systems, 30, 2017

    2017

  75. [83]

    Multi-label learning with weak label

    Yu-Yin Sun, Yin Zhang, and Zhi-Hua Zhou. Multi-label learning with weak label. In AAAI conference on artificial intelligence, volume 24, pages 593–598, 2010

    2010

  76. [84]

    Mean teachers are better role models: Weight-averaged consistency targets improve semi-supervised deep learning results.Advances in neural information processing systems, 30, 2017

    Antti Tarvainen and Harri Valpola. Mean teachers are better role models: Weight-averaged consistency targets improve semi-supervised deep learning results.Advances in neural information processing systems, 30, 2017

    2017

  77. [85]

    Classification of acoustic events using svm-based clustering schemes

    Andrey Temko and Climent Nadeu. Classification of acoustic events using svm-based clustering schemes. Pattern Recognition, 39(4):682–694, 2006

    2006

  78. [86]

    Perspectives in machine learning for wildlife conservation

    Devis Tuia, Benjamin Kellenberger, Sara Beery, Blair R Costelloe, Silvia Zuffi, Benjamin Risse, Alexander Mathis, Mackenzie W Mathis, Frank Van Langevelde, Tilo Burghardt, et al. Perspectives in machine learning for wildlife conservation. Nature communications, 13(1):1–15, 2022. 14

    2022

  79. [87]

    Building a bird recognition app and large scale dataset with citizen scientists: The fine print in fine-grained dataset collection

    Grant Van Horn, Steve Branson, Ryan Farrell, Scott Haber, Jessie Barry, Panos Ipeirotis, Pietro Perona, and Serge Belongie. Building a bird recognition app and large scale dataset with citizen scientists: The fine print in fine-grained dataset collection. In Proceedings of the...

    2015

  80. [88]

    Benchmarking representation learning for natural world image collections

    Grant Van Horn, Elijah Cole, Sara Beery, Kimberly Wilber, Serge Belongie, and Oisin Mac Aodha. Benchmarking representation learning for natural world image collections. In IEEE/CVF Conference on Computer Vision and Pattern Recognition, pages 12884–12893, 2021

    2021

  81. [89]

    Exploring fine-grained audiovisual categorization with the SSW60 dataset

    Grant Van Horn, Rui Qian, Kimberly Wilber, Hartwig Adam, Oisin Mac Aodha, and Serge Belongie. Exploring fine-grained audiovisual categorization with the SSW60 dataset. In European Conference on Computer Vision, pages 271–289. Springer, 2022

    2022

  82. [90]

    Understanding objects in detail with fine-grained attributes

    Andrea Vedaldi, Siddharth Mahendran, Stavros Tsogkas, Subhransu Maji, Ross Girshick, Juho Kannala, Esa Rahtu, Iasonas Kokkinos, Matthew B Blaschko, David Weiss, et al. Understanding objects in detail with fine-grained attributes. In Proceedings of the IEEE conference on comput...

    2014

  83. [91]

    A collection of fully-annotated soundscape recordings from neotropical coffee farms in colombia and costa rica

    Alvaro Vega-Hidalgo, Stefan Kahl, Viviana Symes, Ruiz-Gutiérrez, Ingrid Molina-Mora, Fernando Cediel, Luis Sandoval, and Holger Klinck. A collection of fully-annotated soundscape recordings from neotropical coffee farms in colombia and costa rica. URL https: // doi. org/ 10. 5...

  84. [92]

    Audio transformers: Transformer architectures for large scale audio understanding

    Prateek Verma and Jonathan Berger. Audio transformers: Transformer architectures for large scale audio understanding. adieu convolutions. arXiv preprint arXiv:2105.00335, 2021

    2021 arXiv

  85. [93]

    Towards automatic bird detection: An annotated and segmented acoustic dataset of seven picidae species

    Ester Vidaña-Vila, Joan Navarro, and Rosa Ma Alsina-Pagès. Towards automatic bird detection: An annotated and segmented acoustic dataset of seven picidae species. Data, 2(2):18, 2017

    2017

  86. [94]

    C. Wah, S. Branson, P. Welinder, P. Perona, and S. Belongie. The Caltech-UCSD Birds-200-2011 Dataset. Technical Report CNS-TR-2011-001, California Institute of Technology, 2011

    2011

  87. [95]

    Early detection of rapid barred owl population growth within the range of the california spotted owl advises the precautionary principle

    Connor M Wood, RJ Gutiérrez, John J Keane, and M Zachariah Peery. Early detection of rapid barred owl population growth within the range of the california spotted owl advises the precautionary principle. The Condor, 122(1):duz058, 2020

    2020

  88. [96]

    https://xeno-canto.org, 2024

    Xeno-Canto. https://xeno-canto.org, 2024. Accessed on Jun 1, 2024

    2024

  89. [97]

    Zero-shot learning—a compre- hensive evaluation of the good, the bad and the ugly

    Yongqin Xian, Christoph H Lampert, Bernt Schiele, and Zeynep Akata. Zero-shot learning—a compre- hensive evaluation of the good, the bad and the ugly. IEEE transactions on pattern analysis and machine intelligence, 41(9):2251–2265, 2018

    2018

  90. [98]

    Large-scale weakly supervised audio classification using gated convolutional neural network

    Yong Xu, Qiuqiang Kong, Wenwu Wang, and Mark D Plumbley. Large-scale weakly supervised audio classification using gated convolutional neural network. In IEEE international conference on acoustics, speech and signal processing (ICASSP), pages 121–125, 2018

    2018

  91. [99]

    Automated classification of avian vocal activity using acoustic indices in regional and heterogeneous datasets

    Daniel A Yip, C Lisa Mahon, Alexander G MacPhail, and Erin M Bayne. Automated classification of avian vocal activity using acoustic indices in regional and heterogeneous datasets. Methods in Ecology and Evolution, 12(4):707–719, 2021

    2021

  92. [100]

    Mosi: multimodal corpus of sentiment intensity and subjectivity analysis in online opinion videos

    Amir Zadeh, Rowan Zellers, Eli Pincus, and Louis-Philippe Morency. Mosi: multimodal corpus of sentiment intensity and subjectivity analysis in online opinion videos. arXiv preprint arXiv:1606.06259, 2016

    2016 arXiv

  93. [101]

    Multimodal language analysis in the wild: Cmu-mosei dataset and interpretable dynamic fusion graph

    AmirAli Bagher Zadeh, Paul Pu Liang, Soujanya Poria, Erik Cambria, and Louis-Philippe Morency. Multimodal language analysis in the wild: Cmu-mosei dataset and interpretable dynamic fusion graph. In Association for Computational Linguistics (ACL), pages 2236–2246, 2018

    2018

  94. [102]

    A survey of audio classification using deep learning

    Khalid Zaman, Melike Sah, Cem Direkoglu, and Masashi Unoki. A survey of audio classification using deep learning. IEEE Access, 2023

    2023

  95. [103]

    Acoustic identification of mexican bats based on taxonomic and ecological constraints on call design

    Veronica Zamora-Gutierrez, Celia Lopez-Gonzalez, M Cristina MacSwiney Gonzalez, Brock Fenton, Gareth Jones, Elisabeth KV Kalko, Sebastien J Puechmaille, Vassilios Stathopoulos, and Kate E Jones. Acoustic identification of mexican bats based on taxonomic and ecological constrai...

    2016

  96. [104]

    mixup: Beyond empirical risk minimization

    Hongyi Zhang, Moustapha Cisse, Yann N Dauphin, and David Lopez-Paz. mixup: Beyond empirical risk minimization. In International Conference on Learning Representations, 2018

    2018

  97. [105]

    Acknowledging the unknown for multi-label learning with single positive labels

    Donghao Zhou, Pengfei Chen, Qiong Wang, Guangyong Chen, and Pheng-Ann Heng. Acknowledging the unknown for multi-label learning with single positive labels. In European Conference on Computer Vision, pages 423–440. Springer, 2022. 15 A Appendix Dataset documentation, licensing ...

    2022

  98. [106]

    For all authors... (a) Do the main claims made in the abstract and introduction accurately reflect the paper’s contributions and scope? [Yes] (b) Did you describe the limitations of your work? [Yes] See Sec. 6. (c) Did you discuss any potential negative societal impacts of you...

  99. [107]

    (a) Did you state the full set of assumptions of all theoretical results? [N/A] (b) Did you include complete proofs of all theoretical results? [N/A]

    If you are including theoretical results... (a) Did you state the full set of assumptions of all theoretical results? [N/A] (b) Did you include complete proofs of all theoretical results? [N/A]

  100. [108]

    for benchmarks)

    If you ran experiments (e.g. for benchmarks)... (a) Did you include the code, data, and instructions needed to reproduce the main experi- mental results (either in the supplemental material or as a URL)? [Yes] See Sec. 4. The URL for iNatSounds is provided in the Sec. 1. (b) D...

  101. [109]

    (a) If your work uses existing assets, did you cite the creators? [Yes] We use assets from iNaturalist, See Section 3

    If you are using existing assets (e.g., code, data, models) or curating/releasing new assets... (a) If your work uses existing assets, did you cite the creators? [Yes] We use assets from iNaturalist, See Section 3. (b) Did you mention the license of the assets? [Yes] We includ...

  102. [110]

    If you used crowdsourcing or conducted research with human subjects... (a) Did you include the full text of instructions given to participants and screenshots, if applicable? [N/A] (b) Did you describe any potential participant risks, with links to Institutional Review Board (...

Pith tools

Reviewed August 7, 2026 · model on record in the stance chip above.