vIR-OLO: an open-source YOLO-based interface for functionalgroup detection and localization in infrared spectra

<jats:p>Accurate identification of functional groups from infrared (IR) spectra is a fundamental task in molecular characterization. Recent data-driven approaches have demonstrated strong performance in functional group recognition; however, most methods formulate the problem as a classification task and do not explicitly identify the spectral regions responsible for their predictions. This limits their utility for annotation workflows, model verification, and human-guided interpretation. Here, we introduce vIR-OLO , an open-source framework that reformulates functional group identification in IR spectroscopy as an object detection problem using the YOLO architecture. Rather than predicting only the presence of functional groups, vIR-OLO localizes characteristic spectral features through bounding-box annotations directly on IR spectra. We developed and benchmarked three pretrained models based on alternative functional group sets containing 10, 12, and 13 classes, respectively. Across all configurations, the models achieved strong incidence-based detection performance (micro-averaged F1 scores exceeding 0.84) while maintaining accurate localization of matched spectral regions (mean Intersection over Union greater than 0.80). vIR-OLO is available at https://github.com/ChemAI-Lab/vIR-OLO, and the pre-trained models are available at https://huggingface.co/ChemAI-Lab.</jats:p>