A Public Image Dataset for Surface Defect Detection of Water-Based Coated Wood Products

We present an image dataset of surface defects of waterborne paint sprayed on wood products, which includes four types of defects: scratches, cracks, bubbles, and holes. The dataset has a total of 13,400 images of different samples. The data collection process is from the actual production workshop of Baijiawang Co., Ltd. in Jiangshan City, China, and is captured using professional high-resolution industrial cameras, and the images are passed through a professional to identify the defect categories. The data was collected to support deep learning inspections deployed on the actual assembly line, where researchers and developers can use the dataset to build machine learning solutions that automatically detect different types of defects in the images and take timely and remedial action on detected defects, thus safeguarding the integrity of the finish of wood products.