Preprocessed Yarn-Breakage Image Dataset from Robot-Assisted Ring-Spinning Inspection (Textile Manufacturing, 2024)

The Preprocessed images are one-to-one enhanced counterparts of the Raw images and are designed to improve yarn and fiber feature visibility for yarn-breakage classification and detection. They retain the same filenames, class-folder organization, image resolution, spatial geometry, and three-position yarn-state labels as the corresponding Raw images.Preprocessing & PairingThe Preprocessed images were generated from the Raw images using a fixed enhancement pipeline including Prewitt-based edge extraction, edge-feature fusion, linear brightness adjustment, pixel-value clipping, and gamma correction.

The preprocessing pipeline did not crop, resize, rotate, or geometrically transform the images. Therefore, the Preprocessed images should not be counted as additional independent original samples, but as one-to-one enhanced versions of the Raw image records.Temporal metadata were derived from the timestamped Raw BMP filenames, which serve as the canonical original acquisition records. Because the Preprocessed images are paired one-to-one with the Raw images using identical filenames and class-folder organization, the temporal metadata derived from the Raw BMP image records can also be linked to the corresponding Preprocessed records.Metadata DetailsA shared common_metadata/ folder is provided for both the Raw and Preprocessed records.