Supervised Detection of Strongly-Labelled Whale Calls

Results

Task description

Supervised sound event detection of 7 different call types from the blue whale and the fin whale. Models must handle calls that occur only ~6% of the time and generalise across PAM recordings from different time periods and sites around Antarctica with highly variable soundscapes. Submissions are evaluated using macro F1-score (IoU ≥ 0.3) on four held-out recording sites.

Task description page

Awards

The official winning submission is determined by the official evaluation ranking. The Jury Award is selected by the task organisers based on the submitted systems and technical reports.

Official winning submission

Ye, Tong and Liu, Hongning and Xiao, Feiyang and Zhu, Qiaoxi and Guan, Jian
Group of Intelligent Signal Processing, Harbin Engineering University
GISP@HEU's submission for task 2: a consensus-ensemble YOLO system for temporal detection of Antarctic whale calls

Jury award

Michael Moshe Michelashvili, Danielle Hausler, Amit Galor, Shai Nahum Gefen, Tomer Nachshon, Yuval Mendelson and Naama Yochai
Deep Voice Foundation
Drawing boxes in the deep: spectrogram object detection for Antarctic whale calls
They have demonstrated a substantial improvement in the performance of their models since last year, achieving a second-place ranking in this edition. They're also the only team this year to open-source their code on GitHub.

Submission ranking

Submissions are ranked on the overall macro F1-score (IoU ≥ 0.3) across the four evaluation sites. Precision and recall are reported as supplementary metrics.

Submission information Rank Metrics
Corresponding Author System name Technical
report		 Official
rank		 F-score
(%)		 Precision
(%)		 Recall
(%)
Yongyi Deng Deng_WHU_1 Deng_WHU 8 46.6 50.8 43.0
Yongyi Deng Deng_WHU_2 Deng_WHU 9 46.0 47.5 44.7
Gherasim, George-Daniel Gherasim_ENSTA_1 Gherasim_ENSTA 10 45.5 53.9 39.4
Gherasim, George-Daniel Gherasim_ENSTA_2 Gherasim_ENSTA 5 47.6 56.0 41.4
Jian Guan Guan_GISP_HEU_1 Guan_GISP_HEU 1 49.9 47.5 52.5
Jian Guan Guan_GISP_HEU_2 Guan_GISP_HEU 17 39.2 51.6 31.6
Jian Guan Guan_GISP_HEU_3 Guan_GISP_HEU 24 34.6 41.7 29.6
Jian Guan Guan_GISP_HEU_4 Guan_GISP_HEU 31 7.4 14.7 4.9
Isoda, Daito Isoda_UT Isoda_UT 22 35.6 34.7 36.6
Hugo Magaldi Magaldi_MNHN Magaldi_MNHN 26 27.1 26.5 27.6
Marolt, Matija and Bones, Eva Marolt_UL_1 Marolt_UL 18 38.4 54.7 29.6
Marolt, Matija and Bones, Eva Marolt_UL_2 Marolt_UL 7 46.7 51.2 43.0
Marolt, Matija and Bones, Eva Marolt_UL_3 Marolt_UL 13 40.5 55.7 31.8
Marolt, Matija and Bones, Eva Marolt_UL_4 Marolt_UL 30 15.4 27.7 10.6
Matthias Nagl Nagl_JKU Nagl_JKU 16 39.3 54.4 30.7
Matthias Nagl Nagl_JKU_NAVE Nagl_JKU_NAVE 21 36.1 35.9 36.4
Ragib Amin Nihal Nihal_ScienceTokyo_1 Nihal_ScienceTokyo 14 39.9 54.9 31.3
Ragib Amin Nihal Nihal_ScienceTokyo_2 Nihal_ScienceTokyo 11 43.3 51.7 37.2
Ragib Amin Nihal Nihal_ScienceTokyo_3 Nihal_ScienceTokyo 15 39.4 55.0 30.7
Ragib Amin Nihal Nihal_ScienceTokyo_4 Nihal_ScienceTokyo 23 35.1 56.6 25.4
Liz Ferguson Ferguson_OSA Ferguson_OSA 28 22.7 57.3 14.2
Saeed, Maaz SMT_IDMT_1 Saeed_SMT_IDMT 25 27.6 25.2 30.5
Saeed, Maaz SMT_IDMT_2 Saeed_SMT_IDMT 20 37.6 30.6 48.6
Saeed, Maaz SMT_IDMT_3 Saeed_SMT_IDMT 29 16.6 11.2 32.0
Vogler, Marie Vogler_AWI Vogler_AWI 27 22.9 16.9 35.6
Michelashvili, Michael Moshe Yochai_DV_1 Yochai_DV 6 46.9 51.7 42.8
Michelashvili, Michael Moshe Yochai_DV_2 Yochai_DV 3 48.5 53.3 44.5
Michelashvili, Michael Moshe Yochai_DV_3 Yochai_DV 2 48.6 51.1 46.4
Michelashvili, Michael Moshe Yochai_DV_4 Yochai_DV 12 42.1 59.7 32.5
Xixin Zhang Zhang_SDSU Zhang_SDSU 4 48.3 52.0 45.1
Liz Ferguson yolo_baseline Ferguson2025 19 37.6 38.4 36.8

Subsets

Submission ranking across the four evaluation sites: DDU2021, Kerguelen2020, DDU2019, Kerguelen2019. Showing the macro F1-score (%).

Submission information Rank Sites (F-score %)
Corresponding Author System name Technical
report		 Official
rank		 F-score
(%)		 DDU2021 Kerguelen2020 DDU2019 Kerguelen2019
Yongyi Deng Deng_WHU_1 Deng_WHU 8 46.6 51.4 43.1 44.0 34.6
Yongyi Deng Deng_WHU_2 Deng_WHU 9 46.0 51.5 39.2 47.0 34.8
Gherasim, George-Daniel Gherasim_ENSTA_1 Gherasim_ENSTA 10 45.5 47.4 45.9 43.2 36.0
Gherasim, George-Daniel Gherasim_ENSTA_2 Gherasim_ENSTA 5 47.6 49.3 50.0 45.1 38.6
Jian Guan Guan_GISP_HEU_1 Guan_GISP_HEU 1 49.9 50.4 40.8 47.9 34.2
Jian Guan Guan_GISP_HEU_2 Guan_GISP_HEU 17 39.2 40.9 55.4 41.5 29.4
Jian Guan Guan_GISP_HEU_3 Guan_GISP_HEU 24 34.6 nan 40.4 47.1 34.9
Jian Guan Guan_GISP_HEU_4 Guan_GISP_HEU 31 7.4 nan 37.4 0.1 1.0
Isoda, Daito Isoda_UT Isoda_UT 22 35.6 34.7 38.6 33.0 25.8
Hugo Magaldi Magaldi_MNHN Magaldi_MNHN 26 27.1 21.6 29.0 33.3 30.0
Marolt, Matija and Bones, Eva Marolt_UL_1 Marolt_UL 18 38.4 38.5 50.0 37.4 28.9
Marolt, Matija and Bones, Eva Marolt_UL_2 Marolt_UL 7 46.7 48.8 46.3 47.0 39.4
Marolt, Matija and Bones, Eva Marolt_UL_3 Marolt_UL 13 40.5 42.0 51.5 38.1 27.3
Marolt, Matija and Bones, Eva Marolt_UL_4 Marolt_UL 30 15.4 13.6 23.6 14.1 10.9
Matthias Nagl Nagl_JKU Nagl_JKU 16 39.3 39.8 47.9 36.3 33.7
Matthias Nagl Nagl_JKU_NAVE Nagl_JKU_NAVE 21 36.1 30.1 34.4 30.6 30.2
Ragib Amin Nihal Nihal_ScienceTokyo_1 Nihal_ScienceTokyo 14 39.9 42.5 52.4 31.5 24.9
Ragib Amin Nihal Nihal_ScienceTokyo_2 Nihal_ScienceTokyo 11 43.3 45.8 49.9 37.3 29.1
Ragib Amin Nihal Nihal_ScienceTokyo_3 Nihal_ScienceTokyo 15 39.4 42.4 52.6 30.5 24.2
Ragib Amin Nihal Nihal_ScienceTokyo_4 Nihal_ScienceTokyo 23 35.1 38.3 52.3 25.2 20.2
Liz Ferguson Ferguson_OSA Ferguson_OSA 28 22.7 16.9 32.9 11.1 14.7
Saeed, Maaz SMT_IDMT_1 Saeed_SMT_IDMT 25 27.6 50.9 34.7 0.6 1.4
Saeed, Maaz SMT_IDMT_2 Saeed_SMT_IDMT 20 37.6 43.6 26.5 31.0 30.5
Saeed, Maaz SMT_IDMT_3 Saeed_SMT_IDMT 29 16.6 32.1 16.3 1.2 1.3
Vogler, Marie Vogler_AWI Vogler_AWI 27 22.9 22.1 21.9 16.0 16.0
Michelashvili, Michael Moshe Yochai_DV_1 Yochai_DV 6 46.9 52.4 45.3 50.1 35.1
Michelashvili, Michael Moshe Yochai_DV_2 Yochai_DV 3 48.5 54.7 45.9 44.4 30.2
Michelashvili, Michael Moshe Yochai_DV_3 Yochai_DV 2 48.6 54.8 43.6 44.8 30.9
Michelashvili, Michael Moshe Yochai_DV_4 Yochai_DV 12 42.1 42.4 46.0 31.0 23.5
Xixin Zhang Zhang_SDSU Zhang_SDSU 4 48.3 50.2 46.0 36.0 35.1
Liz Ferguson yolo_baseline Ferguson2025 19 37.6 31.2 32.0 36.0 30.0

Technical reports

Class-aware heterogeneous spectrogram detectors for biodcase 2026 whale call event detection

Deng, Yongyi
Wuhan University, Tampere University, Northwestern Polytechnical University, Waseda University

Baleen Whale Deep Learning Detection and Classification Network

Ferguson, Liz and Alongi, Gabriela
Ocean Science Analytics

A three-channel physics-informed TCN for cross-site Antarctic blue and fin whale call detection

Gherasim, George-Daniel
ENSTA, Institut Polytechnique de Paris

GISP@HEU’s submission for task 2: a consensus-ensemble yolo system for temporal detection of antarctic whale calls

Ye, Tong and Liu, Hongning and Xiao, Feiyang and Zhu, Qiaoxi and Guan, Jian
Group of Intelligent Signal Processing, Harbin Engineering University

Joint whale call detection and recording location classification using multitask learning for biodcase challnege 2026 task 2

Isoda, Daito and Arita, Ryoko and Okamoto, Yuki and Tonami, Noriyuki and Saito, Yuki and Saruwatari, Hiroshi
University of Tokyo, NEC Corporation

SwinWaveD: multi-resolution whale-call detection with calibration and wavelet-ridge verification

Magaldi, Hugo
Muséum National d’Histoire Naturelle

A multi-resolution HTS-AT transformer with feature-pyramid fusion for Antarctic blue and fin whale call detection

Marolt, Matija and Bones, Eva
University of Ljubljana, Faculty of Computer and Information Science

Hard-negative mining, mean-teacher, and class-specialist ensembling for antarctic baleen whale call detection

Matthias Nagl
Johannes Kepler University Linz

NAVE: a normalized, adaptive conformer for antarctic whale vocalization-event detection

Matthias Nagl
Johannes Kepler University Linz

Whale call detection with a frequency-dynamic CRNN ensemble

Ragib Amin Nihal
Institute of Science Tokyo

Comparison of vision-based object detectors and an audio transformer for whale sound event detection

Saeed, Maaz and Bidarouni, Amir Latifi and Lukashevich, Hanna and Abeßer, Jakob
Semantic Media Technologies, Fraunhofer IDMT

YOLOv11-Based Detection of Blue Whale and Fin Whale Vocalizations via Spectrogram Object Detection

Vogler, Marie and Savchik, Alex
University of Hamburg, Alfred Wegener Institute of Polar and Marine Research

Drawing boxes in the deep: spectrogram object detection for Antarctic whale calls

Michelashvili, Michael Moshe and Hausler, Danielle and Galor, Amit and Gefen, Shai Nahum and Nachshon, Tomer and Mendelson, Yuval and Yochai, Naama
Deep Voice Foundation

Transformer object detection on spectrograms for antarctic blue and fin whale call detection

Xixin Zhang
San Diego State University