AutoFish: Dataset and Benchmark for Fine-grained Analysis of Fish

Stefan Hein Bengtson; Daniel Lehotský; Vasiliki Ismiroglou; Niels Madsen; Thomas B. Moeslund; Malte Pedersen

AutoFish: Dataset and Benchmark for Fine-grained Analysis of Fish

Stefan Hein Bengtson^*, Daniel Lehotský, Vasiliki Ismiroglou, Niels Madsen, Thomas B. Moeslund, Malte Pedersen

^*Corresponding author for this work

Research output: Contribution to book/anthology/report/conference proceeding › Article in proceeding › Research › peer-review

Abstract

Automated fish documentation processes are in the near future expected to play an essential role in sustainable fisheries management and for addressing challenges of overfishing. In this paper, we present a novel and publicly available dataset named AutoFish designed for fine-grained fish analysis. The dataset comprises 1,500 images of 454 specimens of visually similar fish placed in various constellations on a white conveyor belt and annotated with instance segmentation masks, IDs, and length measurements. The data was collected in a controlled environment using an RGB camera. The annotation procedure involved manual point annotations, initial segmentation masks proposed by the Segment Anything Model (SAM), and subsequent manual correction of the masks. We establish baseline instance segmentation results using two variations of the Mask2Former architecture, with the best performing model reaching an mAP of 89.15%. Additionally, we present two baseline length estimation methods, the best performing being a custom MobileNetV2-based regression model reaching an MAE of 0.62cm in images with no occlusion and 1.38cm in images with occlusion. Link to project page: https://vap.aau.dk/autofish/

Original language	English
Title of host publication	Proceedings - 2025 IEEE Winter Conference on Applications of Computer Vision Workshops, WACVW 2025
Publication status	Accepted/In press - 2025

Series	IEEE/CVF Winter Conference on Applications of Computer Vision Workshops (WACVW)
ISSN	2572-4398

Bibliographical note

Fish used in these experiments were caught and landed by fishermen following relevant legislation and normal fishing procedures. The Danish Ministry of Food, Agriculture and Fisheries of Denmark was contacted before fish collection to ensure compliance with legislation. The fish were dead at landing and only dead fish were included in this experiment. There is no conflict with the European Union (EU) directive on animal experimentation (article 3, 20.10.2010, Official Journal of the European Union L276/39) and Danish law (BEK nr 12, 07/01/2016). The laboratory facilities used at Aalborg University are approved according to relevant legislation.

UN SDGs

This output contributes to the following UN Sustainable Development Goals (SDGs)

Access to Document

AutoFish_MACVI_WACV25_vbnAccepted author manuscript, 6.45 MBLicence: CC BY 4.0

https://arxiv.org/abs/2501.03767Licence: CC BY 4.0

AutoFish
Bengtson, S. H. (Creator), Lehotský, D. (Creator), Ismiroglou, V. (Creator), Madsen, N. (Creator), Moeslund, T. B. (Creator) & Pedersen, M. (Creator), Hugging Face, 15 Oct 2024
DOI: 10.57967/hf/3990, https://huggingface.co/datasets/vapaau/autofish
Dataset

Cite this

@inproceedings{7e071466f86d4027ae43a509b894762c,

title = "AutoFish: Dataset and Benchmark for Fine-grained Analysis of Fish",

abstract = "Automated fish documentation processes are in the near future expected to play an essential role in sustainable fisheries management and for addressing challenges of overfishing. In this paper, we present a novel and publicly available dataset named AutoFish designed for fine-grained fish analysis. The dataset comprises 1,500 images of 454 specimens of visually similar fish placed in various constellations on a white conveyor belt and annotated with instance segmentation masks, IDs, and length measurements. The data was collected in a controlled environment using an RGB camera. The annotation procedure involved manual point annotations, initial segmentation masks proposed by the Segment Anything Model (SAM), and subsequent manual correction of the masks. We establish baseline instance segmentation results using two variations of the Mask2Former architecture, with the best performing model reaching an mAP of 89.15%. Additionally, we present two baseline length estimation methods, the best performing being a custom MobileNetV2-based regression model reaching an MAE of 0.62cm in images with no occlusion and 1.38cm in images with occlusion. Link to project page: https://vap.aau.dk/autofish/",

author = "Bengtson, {Stefan Hein} and Daniel Lehotsk{\'y} and Vasiliki Ismiroglou and Niels Madsen and Moeslund, {Thomas B.} and Malte Pedersen",

note = "Fish used in these experiments were caught and landed by fishermen following relevant legislation and normal fishing procedures. The Danish Ministry of Food, Agriculture and Fisheries of Denmark was contacted before fish collection to ensure compliance with legislation. The fish were dead at landing and only dead fish were included in this experiment. There is no conflict with the European Union (EU) directive on animal experimentation (article 3, 20.10.2010, Official Journal of the European Union L276/39) and Danish law (BEK nr 12, 07/01/2016). The laboratory facilities used at Aalborg University are approved according to relevant legislation.",

year = "2025",

language = "English",

series = "IEEE/CVF Winter Conference on Applications of Computer Vision Workshops (WACVW)",

publisher = "IEEE (Institute of Electrical and Electronics Engineers)",

booktitle = "Proceedings - 2025 IEEE Winter Conference on Applications of Computer Vision Workshops, WACVW 2025",

}

Bengtson, SH , Lehotský, D , Ismiroglou, V , Madsen, N , Moeslund, TB & Pedersen, M 2025, AutoFish: Dataset and Benchmark for Fine-grained Analysis of Fish. in Proceedings - 2025 IEEE Winter Conference on Applications of Computer Vision Workshops, WACVW 2025. IEEE/CVF Winter Conference on Applications of Computer Vision Workshops (WACVW). <https://arxiv.org/abs/2501.03767>

AutoFish: Dataset and Benchmark for Fine-grained Analysis of Fish. / Bengtson, Stefan Hein ; Lehotský, Daniel ; Ismiroglou, Vasiliki et al.
Proceedings - 2025 IEEE Winter Conference on Applications of Computer Vision Workshops, WACVW 2025. 2025. (IEEE/CVF Winter Conference on Applications of Computer Vision Workshops (WACVW)).

Research output: Contribution to book/anthology/report/conference proceeding › Article in proceeding › Research › peer-review

TY - GEN

T1 - AutoFish: Dataset and Benchmark for Fine-grained Analysis of Fish

AU - Bengtson, Stefan Hein

AU - Lehotský, Daniel

AU - Ismiroglou, Vasiliki

AU - Madsen, Niels

AU - Moeslund, Thomas B.

AU - Pedersen, Malte

N1 - Fish used in these experiments were caught and landed by fishermen following relevant legislation and normal fishing procedures. The Danish Ministry of Food, Agriculture and Fisheries of Denmark was contacted before fish collection to ensure compliance with legislation. The fish were dead at landing and only dead fish were included in this experiment. There is no conflict with the European Union (EU) directive on animal experimentation (article 3, 20.10.2010, Official Journal of the European Union L276/39) and Danish law (BEK nr 12, 07/01/2016). The laboratory facilities used at Aalborg University are approved according to relevant legislation.

PY - 2025

Y1 - 2025

N2 - Automated fish documentation processes are in the near future expected to play an essential role in sustainable fisheries management and for addressing challenges of overfishing. In this paper, we present a novel and publicly available dataset named AutoFish designed for fine-grained fish analysis. The dataset comprises 1,500 images of 454 specimens of visually similar fish placed in various constellations on a white conveyor belt and annotated with instance segmentation masks, IDs, and length measurements. The data was collected in a controlled environment using an RGB camera. The annotation procedure involved manual point annotations, initial segmentation masks proposed by the Segment Anything Model (SAM), and subsequent manual correction of the masks. We establish baseline instance segmentation results using two variations of the Mask2Former architecture, with the best performing model reaching an mAP of 89.15%. Additionally, we present two baseline length estimation methods, the best performing being a custom MobileNetV2-based regression model reaching an MAE of 0.62cm in images with no occlusion and 1.38cm in images with occlusion. Link to project page: https://vap.aau.dk/autofish/

AB - Automated fish documentation processes are in the near future expected to play an essential role in sustainable fisheries management and for addressing challenges of overfishing. In this paper, we present a novel and publicly available dataset named AutoFish designed for fine-grained fish analysis. The dataset comprises 1,500 images of 454 specimens of visually similar fish placed in various constellations on a white conveyor belt and annotated with instance segmentation masks, IDs, and length measurements. The data was collected in a controlled environment using an RGB camera. The annotation procedure involved manual point annotations, initial segmentation masks proposed by the Segment Anything Model (SAM), and subsequent manual correction of the masks. We establish baseline instance segmentation results using two variations of the Mask2Former architecture, with the best performing model reaching an mAP of 89.15%. Additionally, we present two baseline length estimation methods, the best performing being a custom MobileNetV2-based regression model reaching an MAE of 0.62cm in images with no occlusion and 1.38cm in images with occlusion. Link to project page: https://vap.aau.dk/autofish/

M3 - Article in proceeding

T3 - IEEE/CVF Winter Conference on Applications of Computer Vision Workshops (WACVW)

BT - Proceedings - 2025 IEEE Winter Conference on Applications of Computer Vision Workshops, WACVW 2025

ER -

AutoFish: Dataset and Benchmark for Fine-grained Analysis of Fish

Abstract

Bibliographical note

UN SDGs

Access to Document

Fingerprint

Datasets

AutoFish

Cite this