Cornell Researchers Released WildFin Behavior Dataset

The newly released dataset features nine hours of annotated fish behavior footage recorded in Curaçao.

Updated on Oct. 1, 2026 in Aquariums

Cornell Researchers Released WildFin Behavior Dataset

Live Poll

Do you believe using artificial intelligence to track wildlife behavior is beneficial for ecosystem conservation?

Cornell University researchers created the WildFin dataset to improve computer vision analysis of fish behavior in the wild. The study was presented on September 8, 2026, at the European Conference on Computer Vision.

Why it matters

Existing computer vision models often struggle with wildlife footage due to a lack of specialized training data. This dataset provides the necessary tools for scientists to better monitor ecosystem health and analyze complex aquatic behaviors.

The WildFin dataset includes nine hours of video collected over 1,400 hours of field research. The team spent 600 hours manually annotating fish behaviors, such as foraging and fighting, to train recognition models.

The players

Cornell University

This is a private Ivy League research university based in Ithaca, New York, that leads significant scientific and technological studies.

Abigail Grassick

She is a researcher who presented the WildFin findings at the European Conference on Computer Vision in Malmö, Sweden.

National Science Foundation

This is an independent federal agency of the United States that supports fundamental research and education in all the non-medical fields of science and engineering.

NVIDIA

This is a multinational technology company known for designing graphics processing units and hardware that facilitate complex machine learning and AI research.

University of Colorado, Boulder

This is a public research university that contributed annotated aquatic footage to the WildFin project.

The details

The research team combined footage of individual fish tracked by divers with broader recordings of coral reef communities near Curaçao. Supported by the National Science Foundation and NVIDIA, the project also incorporates annotated footage provided by colleagues from the University of Colorado, Boulder.

September 8, 2026: The WildFin study was presented at the European Conference on Computer Vision.

Culture Shift

This project reflects a growing trend in ecological research toward open-source data sharing for artificial intelligence applications. The WildFin dataset marks a departure from proprietary data silos, following the precedent set by presentations at the European Conference on Computer Vision.

The availability of this dataset allows researchers and hobbyists to better understand the daily habits of fish species in coral reef environments. This shift helps observers distinguish between common behaviors like feeding and foraging in their own aquarium setups or field studies.

The takeaway

Access to high-quality annotated video enables the development of better tools for monitoring aquatic biodiversity and behavior. Researchers hope that sharing these visual datasets will encourage a collaborative effort to improve automated wildlife tracking models.