Reference Model - GU Goby 4sp

This is a YOLO8 object detection model trained on video data obtained from the University of Gothenburg. See metadata for a full description.

Metadata description

Model purpose. The model was trained and employed to extract species observations from videos recorded in shallow underwater habitats on the Swedish west coast, with particular focus to determine the abundance of invasive fish (Neogobius melanostomus) in relation to their relatives (Gobius niger) who are native in the area.

Taxonomic scope. The model was trained to identify the following fish and brachyuran crabs to species and /or family level:

Carcinus maenas (European Green Crab), Ctenolabrus rupestris (Goldsinny Wrasse), Gobius niger (Black Goby), Neogobius melanostomus (Round Goby).

Model input. The video material included 1 hour-recordings using baited remote underwater video setup (BRUV).

Model outputs: Detections made by the model are published at the Global Biodiversity Information Facility (GBIF) under the DOI 10.15468/8m29p6 (https://www.gbif.org/dataset/653b41c5-f839-4df9-b6f5-cfb896ac2b52)

Temporal scope. The model was trained and used on movies collected in the years 2023-2024.

Geographic scope. The model was trained on footage from shallow bay environments on the Swedish West coast, at 1-3 meters depth, mainly centred around small marinas. More specifically the footage was collected in freshwater-influenced fjord system (the 8-fjords+ area) containing multiple Natura-2000 sites and a range of habitats including shallow and deep soft and hard bottoms.

Funding: Project funded by Swedish Environmental Protection Agency (Project grants in the handling of alien invasive species 2020, Environmental research fund), grant nr. 2020-00055, to Leon Green.