Fish-Vista: A Multi-Purpose Dataset for Understanding & Identification of Traits from Images

TL;DR

Fish-Vista is a comprehensive, annotated image dataset of over 69,000 fish images designed for species classification, trait identification, and segmentation, advancing AI applications in biodiversity science.

Contribution

The paper introduces Fish-Vista, the first organismal image dataset for visual trait analysis of aquatic species, with a reproducible processing pipeline and benchmarks for multiple computer vision tasks.

Findings

Benchmark results reveal challenges in long-tailed and out-of-distribution generalization.

The dataset enables research in weakly supervised learning and explainable AI.

Fish-Vista facilitates segmentation of small objects in biodiversity images.

Abstract

We introduce Fish-Visual Trait Analysis (Fish-Vista), the first organismal image dataset designed for the analysis of visual traits of aquatic species directly from images using problem formulations in computer vision. Fish-Vista contains 69,126 annotated images spanning 4,154 fish species, curated and organized to serve three downstream tasks of species classification, trait identification, and trait segmentation. Our work makes two key contributions. First, we perform a fully reproducible data processing pipeline to process images sourced from various museum collections. We annotate these images with carefully curated labels from biological databases and manual annotations to create an AI-ready dataset of visual traits, contributing to the advancement of AI in biodiversity science. Second, our proposed downstream tasks offer fertile grounds for novel computer vision research in addressing a variety of challenges such as long-tailed distributions, out-of-distribution generalization, learning with weak labels, explainable AI, and segmenting small objects. We benchmark the performance of several existing methods for our proposed tasks to expose future research opportunities in AI for biodiversity science problems involving visual traits.