AutoSiQ: A Curated Haploid Arabidopsis thaliana Inflorescence Dataset with a Fine-Grained Silique Ontology and a Deep Learning Application for Haploid Fertility Quantification

Data generation pipeline for AutoSiQ: annotation processing, color extraction, and YOLO label formatting. This companion dataset contains the Python and Matlab utilities used to process raw color-annotated Arabidopsis thaliana inflorescence images into YOLOv5-ready bounding boxes and labels. The seven-class ontology supports both the two-class setup (fertile silique, total silique) used in the paper and fine-grained seven-class training. Seven-class ontology: Fertile silique (f) Fertile crack (f_c) Green fertile (g_f) Mature non-fertile (m) Green non-fertile (g) Crack non-fertile (c) Flower (fl) Key scripts: 0_sanatizing.py: Annotation sanitization and cleanup. 1_5_json_csv_yolov5_sanity_check_bbox.py: Bounding-box validation and sanity checks. 2_json_csv_yolov5_new_annotation.py: Configurable class mapping — converts JSON/CSV annotations to YOLOv5 format. Adjust class settings to export either two-class (paper setup) or all seven classes. generate_disttributions.py: Class distribution analysis and reporting. Matlab utilities (createMask.m, make_patches_and_boundingbox_labels.m): Color-based mask extraction and patch/bounding-box generation. Example data: annotated_png/, raw_jpg/: Sample raw and color-annotated images showing the input format. detection_labels.csv, class_distribution.csv: Reference annotation and distribution files. Usage: Run scripts sequentially to convert your color-annotated images to YOLO format with configurable class granularity.