CMYK-Based Synthetic Datasets for Biological Degradation Detection in Historical Documents

These datasets were created to support research on automatic localization of biological degradation in historical documents and were used in the experiments presented in the accompanying publication. It contains synthetic datasets generated for the detection of biological degradation patterns in historical documents using the proposed CMYK-based blending pipeline. Dataset structure Foxing_feces Contains datasets generated from naturally occurring foxing stains and insect-induced degradation. dataset_combo_cmyk – combined dataset consisting of real historical document images and synthetic images generated using the proposed CMYK-based blending method. syn_dataset_cmyk – synthetic dataset generated exclusively using the proposed CMYK-based blending method. Fungi_bacteria Contains synthetic datasets generated from laboratory-prepared microbial degradation samples. fungus_bacteria_cmyk_all_classes_syn – synthetic dataset containing 12 microbial classes (6 bacterial and 6 fungal species) generated using the proposed CMYK-based blending pipeline. Annotation format All datasets are annotated in the YOLO object detection format, where each annotation file contains: <class_id> <x_center> <y_center> <width> <height> with normalized coordinates in the range ([0,1]). For the Fungi_bacteria dataset, the class mapping is: Class ID Species 0 Sphingomonas aquatilis 1 Rhodococcus degradans 2 Micrococcus sp. 3 Curtobacterium flaccumfaciens 4 Epicoccum layuense 5 Gibellulopsis nigrescens 6 Exiguobacterium indicum 7 Arthrobacter sp. 8 Talaromyces purpureogenus 9 Penicillium menoniorum 10 Aspergillus ochraceus 11 Alternaria alternata For Foxing_feces dataset, the class mapping is: Class ID Class 0 Foxing 1 Insect feces