Dataset: CropScope - A multi-scale image and video dataset for reproducible computer vision

CropScope is an open-source, metadata-driven software suite for the reproducible generation, transformation, and validation of large-scale image datasets across multiple domains, including medicine, agriculture, and autonomous driving. Developed in Java, CropScope integrates structured image cropping, metadata generation, and cloud-native data management through an S3-compatible storage layer (S3Service), enabling secure, efficient, and scalable dataset creation directly from local or cloud environments.The suite provides both graphical and batch interfaces, supports formats such as DICOM and NIfTI, and automatically produces FAIR-compliant metadata in JSON and COCO formats to ensure transparency and traceability. Quality control is performed through the integrated Image Verify Tool, which applies SSIM, PSNR, and dHash metrics to detect near-duplicates, achieving 98.7% deduplication precision and reducing curation time by up to 70%.CropScope has been validated across more than 110,000 images from medical, agricultural, and automotive sources, demonstrating reproducibility, provenance tracking, and high data integrity.