BioBench: A Blueprint to Move Beyond ImageNet for Scientific ML Benchmarks
This addresses the need for reliable AI-for-science benchmarks in ecology, offering a template for other domains, though it is incremental as it builds on existing benchmark concepts.
The paper tackles the problem that ImageNet-1K linear-probe accuracy poorly predicts performance on scientific imagery, showing it explains only 34% of variance on ecology tasks and mis-ranks 30% of models above 75% accuracy. It presents BioBench, an open ecology vision benchmark with 9 tasks, 4 taxonomic kingdoms, and 6 acquisition modalities totaling 3.1M images, providing new signal for computer vision in ecology.
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.