GongChePair: A Paired RGB-Edge Benchmark for Gong-Che Symbol Detection in Aged Documents

GongChePair is a paired RGB-edge dataset for quantitative extraction and symbol detection from aged Gong-Che music documents. Each sample contains an RGB manuscript page and a deterministically generated edge counterpart aligned at page level. The dataset supports research on degraded-document analysis, historical music notation recognition, and robust symbol localization under yellowing, low contrast, background contamination, and broken strokes. The release is intended to include benchmark split metadata, annotation files, and documentation describing the paired RGB-edge construction protocol and evaluation setting used in the associated manuscript.