Community Fish Detection Dataset

Images are available in the following cloud storage folders:

• gs://public-datasets-lila/community-fish-detection-dataset-2026/images (GCP)
• s3://us-west-2.opendata.source.coop/agentmorris/lila-wildlife/community-fish-detection-dataset-2026/images (AWS)
• https://lilawildlife.blob.core.windows.net/lila-wildlife/community-fish-detection-dataset-2026/images (Azure)

We recommend downloading images (the whole folder, or a subset of the folder) using gsutil (for GCP), aws s3 (for AWS), or AzCopy (for Azure). For more information about using gsutil, aws s3, or AzCopy, check out our guidelines for accessing images without using giant zipfiles.

If you prefer to download individual images via http, you can. For example, the thumbnail below appears in the metadata as:

This image can be downloaded directly from any of the following URLs (one for each cloud):

Constituent datasets

Brief descriptions and license information for each dataset are included here; for more detailed information and thumbnails from each dataset, see the fish-datasets GitHub repo. The dataset from which each image was drawn is available in the “dataset” field in the .json file.

The Brackish Dataset

This dataset includes 89 short videos (~14,764 extracted frames) from brackish water in Denmark, annotated with boxes on fish, crabs, and other fauna. This dataset is released under CC BY-SA 4.0.

Deep Vision Fish Dataset

This dataset is constructed from images collected using the Deep Vision system during two surveys from 2017 and 2018 that targeted economically important pelagic species. The fish species annotated are blue whiting, Atlantic herring, mesopelagic fish, and Atlantic mackerel. This dataset is licensed under CC-BY 4.0.

DeepFish

This dataset includes ~40k seafloor images from 20 tropical habitats in northern Australia, aimed at benchmarking habitat‑aware fish detection and segmentation. The original dataset contains a mix of classification, segmentation, and counting labels; only images with segmentation masks were used here, resulting in 310 images with 338 annotations. The subset used here is redistributed under the MIT License.

F4K Detection and Tracking

This dataset includes 17 10-minute videos representing a variety of water conditions, with segmentation masks on a subset of frames. This dataset’s license is not explicitly stated.

FathomNet

FathomNet is an image database containing a variety of marine taxa; the subset used here is intended to capture fish. Specifically, this subset includes descendents of Actinopterygii, Sarcopterygii, Chondrichthyes, and Myxini. This dataset is released under CC-BY-ND 4.0, with a provision allowing use for training machine learning models for commercial and non-commercial applications.

FishCLEF 2015

This dataset includes 93 videos annotated with bounding boxes and fish species (14k annotations on 20k frames). This dataset’s license is not explicitly stated.

Tropical freshwater fish in Northern Australia

This dataset includes 44,112 images with 82,904 bounding box annotations for 23 tropical freshwater fish taxa from northern Australia. Images were derived from Remote Underwater Video (RUV) deployments in deep channel and shallow lowland billabongs, Kakadu National Park, Northern Territory Australia. RUV deployments were conducted during the Supervising Scientists annual fish monitoring program in the 2016, 2017, and 2018 recessional flow period (dry season). This dataset is released under CC-BY 4.0.

Marine Detect

This dataset is itself a composite of several other publicly-available datasets; most of the sub-datasets come either from GBIF or from Roboflow Universe. A list of the sub-datasets is available here. Datasets from GBIF are licensed under CC-BY-NC 4.0; the license is not explicitly listed for the other datasets.

MIT Sea Grant River Herring

This dataset was developed from 1,434 video clips from rivers in Massachusetts, USA, comprising a total of ~262k frames, with ~91k bounding boxes annotated on ~60k of those frames (the remaining ~162k frames were reviewed and classified as empty). This dataset is released under CDLA-permissive 1.0.

Puget Sound Nearshore Fish

This dataset contains 77,739 images sampled from video collected on and around shellfish aquaculture farms in an estuary in the Northeast Pacific, in which 67,990 objects (fish and crustaceans) have been annotated on 30,384 images (the remainder have been annotated as “empty”). This dataset is released under CDLA-permissive 1.0.

Project Natick Underwater Video

This auxiliary release from Microsoft’s “Project Natick” underwater data center experiment provides 1,072 RGB images (with box annotations on fish and squid) recorded by cameras fixed on the outside of the data center pod. This dataset is released as part of an MIT-licensed software project, but the dataset license is not stated explicitly.

Roboflow Fish Dataset

This dataset is a collection of 1,350 images and boxes covering 26 fish categories, sourced from the Web. This dataset is released under CC0.

Salmon Computer Vision

This dataset includes 532,000 frames labeled with bounding boxes on 15 species from a weir on the Kitwanga River in BC, Canada. This dataset is licensed under CC-BY-NC-SA 4.0.

Tasmanian Orange Roughy Stereo Image Machine Learning Dataset (TORSI)

This dataset is a collection of annotated stereo image pairs collected by a net-attached Acoustic and Optical System (AOS) during orange roughy (Hoplostethus atlanticus) biomass surveys off the northeast coast of Tasmania, Australia in July 2019. This dataset is released under CC-BY-NC-SA 4.0.

FishTrack23

This dataset includes approximately 850k bounding boxes across 26k tracks in a variety of ocean conditions. This dataset is released under CC-BY 4.0.

AAU Zebrafish Re-Identification Dataset

This dataset includes ~2200 images of zebrafish in a lab environment with bounding box annotations and individual identifications. Six fish were recorded in total, in two sessions with three fish at a time. A 32 x 32 x 32 cm fish tank was used, with the fish constrained to swim within the first 3.5 cm of the tank. This dataset is released under CC BY 4.0.

PomerFish

This dataset includes 14,989 frames from GoPro videos in Pomeranian rivers (Poland, 2015-2024), with 21,267 boxes on freshwater fish such as perch, grayling, brown trout, roach, and rainbow trout. This dataset is released under CC-BY 4.0.

OBSEA

This dataset includes 4,120 images from the OBSEA seafloor observatory (NW Mediterranean, Spain), with 34,728 fish boxes and a few on other animals, over 74 source classes. This dataset is released under CC-BY 4.0.

CoralscapesV2

This dataset includes 2,433 densely annotated frames from diver cameras at Red Sea coral reefs, with 95-class semantic segmentation and 65,738 fish instance masks. This dataset is released under Apache 2.0.

Orange Chromide Pond Fish Detection

This dataset includes 586 underwater images from ponds in South India, with 10,607 boxes on orange chromide and other cichlids. This dataset is released under CC-BY 4.0.

PEBL Seaweed and Shellfish Farm Fauna

This dataset includes 1,035 frames from fixed underwater cameras (PEBL SubCam) on UK seaweed and shellfish farms, with 2,464 bounding boxes on the animals in eight groups: fish, hermit crabs, crabs, jellyfish, small bottom fish such as gobies, dragonets, starfish, flatfish, and gastropods. Frames come from 2020, 2024, and 2026 deployments at two farms in turbid, temperate water, where detectors trained on clearer water find few animals. This dataset is released under CDLA-permissive 1.0.

CoReFish (Caribbean Coral Reef Fish)

This dataset includes 14,580 frames from diver transect videos at five coral reef sites in the US Virgin Islands (2016-2023), with 100,074 expert-drawn fish boxes and tracks. This dataset is released under CC-BY-NC-SA 4.0.

FjordFish

This dataset contains ~6k annotations on ~3k images across 27 classes. Images were collected from baited and unbaited remote underwater video surveys along the Swedish and Norwegian coasts. Images of coastal cod were also collected from a semi-natural mesocosm at the Institute of Marine Research Flødevigen research station. This dataset is released under CC-BY 4.0.

Newfoundland Marine Refuge Fish Classification Dataset (N-MARINE)

This dataset includes 23,936 frames from baited, lit cameras on crab pots at 400m-500m depth off of Newfoundland (Canada), with 23,890 boxes on eight groundfish and whelk species. This dataset is released under the Canadian Open Government Licence.

WIO-ReefFish

This dataset includes 1,000 4K frames from diver video transects on coral reefs in Kenya and Tanzania, with 6,768 boxes on fish of 24 reef fish families. This dataset is released under CC-BY 4.0.

DIGI4ECO Underwater Fish Tracking

This dataset includes 5,849 frames from 4 videos at three European marine observatories (Adriatic Sea, NW Mediterranean, Galway Bay), with 114,207 fish boxes from manually annotated tracks. This dataset is released under CC-BY 4.0.

YOLO-BRUV

This dataset includes 2,640 frames from 19 baited remote underwater videos at offshore wind farms in the Moray Firth (Scotland), with 13,070 fish boxes (mostly flatfish and haddock) and 2,674 non-fish boxes (mostly hermit crabs and starfish). This dataset is released under CC-BY 4.0.

Labelled Mediterranean Fish Dataset (DEEPECOMAR)

This dataset includes 1,516 images of coastal fish in the Balearic Islands (Spain) from a fixed time-lapse camera, diver videos, and a stereo video, with 17,157 fish boxes over 20 species. This dataset is released under CC-BY 4.0.

Seatizen Atlas (instance segmentation subset)

This dataset includes 1,200 coral reef and lagoon images from snorkellers, divers, and surface vehicles in the western Indian Ocean (Reunion, Mayotte, Europa, Mauritius), with instance polygons for fish, sea cucumbers, urchins, turtles, and reef habitats. This dataset is released under CC-BY 4.0.

AFSC Salmon and Pollock Object Detection

This dataset includes 16,922 frames from a camera inside a commercial pollock trawl in the eastern Bering Sea (2019-2020), with 84,786 boxes on walleye pollock and Pacific salmon. This dataset is released under CC0 1.0.

KSO Mussel Farm BRUV

This dataset includes 910 frames from baited remote underwater videos at a blue mussel farm and a control site on the Swedish West coast, with 1,457 fish boxes (gobies, flatfish) and 207 crab boxes. This dataset is released under CC-BY 4.0.

Griffith Seagrass Fish (Luderick Seagrass)

This dataset includes 4,280 frames from remote underwater video in seagrass meadows on the Gold Coast (Queensland, Australia), with 9,429 boxes and masks on luderick and bream. This dataset is released under CC-BY 4.0.

Mediterranean Fish Census (autoUC)

This dataset includes 13,044 frames from diver GoPro transects at eight sites on the French Riviera, with 68,590 boxes on 40 fish taxa plus jellyfish and sea urchins. This dataset is released under CC0 1.0.

EyeSea Underwater Video

This dataset includes 70,195 frames from monitoring video at three water power sites (a tidal turbine in Scotland, a dam fish ladder in Washington and a river turbine in Alaska), with 45,652 fish boxes. This dataset’s license is not explicitly stated.

Seaclear Marine Debris Dataset

This dataset includes 8,610 ROV images of marine litter on the shallow seabed in Croatia and France, with 40 debris, animal, plant, and robot classes; the fish, sea urchin, sea cucumber, and starfish boxes are used. This dataset is released under CC-BY 4.0.

Roboflow Aquarium Dataset

This dataset includes 638 photos taken through the glass of tanks at two US aquariums, with boxes on fish, sharks, stingrays, jellyfish, and starfish (plus penguins and puffins, which are discarded). This dataset is released under CC-BY 4.0.

RUOD-R (re-annotated Real-world Underwater Object Detection)

This dataset includes 13,112 underwater photos from many sources (e.g. diving videos, aquariums, and web photos), with 104,202 fish and 67,937 non-fish boxes from a dense re-annotation of the RUOD benchmark. This dataset’s license is not explicitly stated.

UIIS10K (Underwater Image Instance Segmentation)

This dataset includes ~10k underwater photos (reefs, open water, aquariums, deep-sea ROV video, lakes) gathered from the web and other public datasets, with instance masks for 10 classes, yielding 19,937 fish and 1,399 non-fish boxes. This dataset is released under Apache 2.0.

NorFisk Dataset

This dataset includes 12,514 crops from GoPro video at Norwegian fish farms, each with one box on a farmed salmonid or a wild saithe. This dataset is released under CC0 1.0.

TrashCan 1.0

This dataset includes 7,212 frames from JAMSTEC deep-sea ROV videos (mostly the Sea of Japan), with instance masks on trash, ROVs, plants, and animals; the fish, eel, crab, and starfish masks are used. This dataset is free for academic teaching and research; commercial use requires permission from JAMSTEC.

Tassie BRUV

This dataset includes 1,912 frames from 28 baited remote underwater videos near boat moorings in Tasmania (Australia), with 5,222 boxes on 19 reef and sand fish species. This dataset is released under CC0 1.0.

Posted by Dan Morris.