FabricSpotDefect: An Annotated Dataset for Identifying Spot Defects in Different Fabric Types

Content: This FabricSpotDefect dataset is designed to improve fabric spot defect detection using computer vision. It includes various types of fabrics such as cotton, linen, silk, denim, and both plain and patterned textiles, featuring spot defects like stains, discolorations, oil marks, and more.

Format: The images are in 2D RGB .jpg format of varying sizes.

Original Images: 1,014 raw images with 3,286 spot defect labels.

Augmented Images: After applying six types of augmentation techniques, the dataset expands to 2,300 images with 7,641 labeled spots.

Data Features: The dataset features both original and augmented images organized into “original” and “augmented” folders, each containing subfolders for training, validation, and testing. All images are resized to 416×416 pixels and undergo six augmentation techniques: flipping, rotation, shearing, saturation, brightness, and noise addition.

How data are acquired: The FabricSpotDefect dataset images were collected from everyday home fabrics under natural lighting using three smartphones( Samsung Galaxy Note20, Samsung Galaxy S20 FE, and Samsung Galaxy A53 5G).

Use Case: To develop an AI model that will identify spot defects in fabrics, so that the model could be used to improve the quality control process in textile production.