AnonymousResearchTiger/PHOEBI

PHOEBI（Phase-contrast Optical bEnchmark for Bacterial Identification）是一个用于光学相位对比显微镜中混合细菌培养物开放世界识别的基准和框架。数据集包含两个子集：PHOEBI-6（主要）和PHOEBI-4（遗留），均基于相同的成像流程。PHOEBI-6包含6种细菌物种（Bacillus subtilis、Bacillus thermoamylovorans、Flavobacterium johnsoniae、Klebsiella aerogenes、Myxococcus xanthus、Pseudomonas fluorescens），约120,000张1024×1024×3 JPEG图像；PHOEBI-4包含4种细菌物种，约14,000张图像。图像以1000×总放大倍数（100×油浸，NA 1.25）拍摄。数据集提供随机80/10/10分割和留出组合（LCO）分割（仅PHOEBI-6），用于支持多标签物种存在检测、组合泛化、开放集拒绝和新类发现等任务。数据以CC BY 4.0许可发布，并包含Croissant 1.0元数据。PHOEBI (Phase-contrast Optical bEnchmark for Bacterial Identification) is a benchmark and framework for open-world identification of mixed bacterial cultures from optical phase-contrast microscopy. The dataset consists of two subsets: PHOEBI-6 (primary) and PHOEBI-4 (legacy), both sharing the same imaging pipeline. PHOEBI-6 includes 6 bacterial species (Bacillus subtilis, Bacillus thermoamylovorans, Flavobacterium johnsoniae, Klebsiella aerogenes, Myxococcus xanthus, Pseudomonas fluorescens) with approximately 120,000 1024×1024×3 JPEG images; PHOEBI-4 includes 4 bacterial species with approximately 14,000 images. Images are captured at 1000× total magnification (100× oil-immersion, NA 1.25). The dataset provides random 80/10/10 splits and leave-combinations-out (LCO) splits (PHOEBI-6 only) to support tasks such as multi-label species presence detection, compositional generalization, open-set rejection, and novel-class discovery. It is released under the CC BY 4.0 license and includes Croissant 1.0 metadata.