Custom Dataset Annotations, Generator Weights, and Processed TILDA-Based Experimental Data for Real-Time Textile Defect Inspection: A Lightweight Super-Resolution Augmented Detection Pipeline

This record contains the experimental data package associated with the study Real-Time Textile Defect Inspection: A Lightweight Super-Resolution Augmented Detection Pipeline. The uploaded files include: custom_dataset_annotations.zip: custom-generated annotations and related experimental metadata used in this study. generator_weights.zip: ESRGAN generator checkpoints for both baseline convolution-based and depthwise-separable-convolution-based model variants. tilda_train_test_yolo_tiling.zip: processed TILDA-based experimental data used in this work, including ESRGAN train/test subsets and YOLO tiling data prepared for downstream detection and segmentation experiments. sr_outputs.zip: 4× super-resolved output images generated during the experiments for visual comparison and qualitative assessment. The original public TILDA dataset was obtained from its original public source and is not redistributed here as a raw standalone dataset. This upload provides the processed experimental data structure used in the reported pipeline. The associated code repository is available on GitHub:https://github.com/ahmet-metin/textile-defect-sr-pipeline If you use this dataset, please also cite the related manuscript.