BioBench: A Blueprint to Move Beyond ImageNet for Scientific ML Benchmarks

TL;DR

BioBench introduces a comprehensive ecology-focused vision benchmark with diverse tasks and modalities, addressing the limitations of ImageNet in predicting scientific imagery performance and enabling more reliable AI-for-science evaluations.

Contribution

BioBench provides a unified, multi-task ecology vision benchmark with 9 tasks, 4 taxonomic kingdoms, and 6 modalities, offering a new standard for scientific machine learning evaluation.

Findings

ImageNet accuracy explains only 34% of variance on ecology tasks.

BioBench's diverse tasks better predict scientific imagery performance.

ViT-L models evaluate efficiently on BioBench, enabling scalable benchmarking.

Abstract

ImageNet-1K linear-probe transfer accuracy remains the default proxy for visual representation quality, yet it no longer predicts performance on scientific imagery. Across 46 modern vision model checkpoints, ImageNet top-1 accuracy explains only 34% of variance on ecology tasks and mis-ranks 30% of models above 75% accuracy. We present BioBench, an open ecology vision benchmark that captures what ImageNet misses. BioBench unifies 9 publicly released, application-driven tasks, 4 taxonomic kingdoms, and 6 acquisition modalities (drone RGB, web video, micrographs, in-situ and specimen photos, camera-trap frames), totaling 3.1M images. A single Python API downloads data, fits lightweight classifiers to frozen backbones, and reports class-balanced macro-F1 (plus domain metrics for FishNet and FungiCLEF); ViT-L models evaluate in 6 hours on an A6000 GPU. BioBench provides new signal for computer vision in ecology and a template recipe for building reliable AI-for-science benchmarks in any domain. Code and predictions are available at https://github.com/samuelstevens/biobench and results at https://samuelstevens.me/biobench.