A dataset of defects in water-based paint coatings on wooden products

By strictly simulating the interference conditions in actual construction, including temperature, humidity, paint thickness and operation methods, four typical paint defects (foreign matter adhesion, scratch, bubble and sagging) are artificially reproduced, which can truly restore the morphological characteristics and distribution rules of defects in the wood water-based paint production process.After image collection,Photoshopis used for unified image preprocessing.

All original images are cropped and standardized to a fixed resolution of 640×640 pixels, eliminating the interference of inconsistent resolution and redundant backgrounds on model training. Subsequently, the professional annotation tool LabelImg is adopted for fine manual bounding box annotation of all defect targets. Following standard object detection annotation specifications, the contours and positions of various defects are accurately marked, and the final annotation files are generated in a standard format compatible with YOLO series detection models.

The dataset contains a total of 3234 valid images, with 1243 foreign matter adhesion instances, 825 scratch instances, 3219 bubble instances, and 3630 sagging instances. With sufficient sample size and diverse defect morphologies, it fully covers typical defects in wood water-based paint coating and has excellent sample representativeness.To ensure the scientificity and rigor of model training, validation and testing and avoid data distribution shift, the dataset is randomly divided into training set, validation set and test set at a ratio of 7:2:1.

The random sampling method ensures the uniform distribution of four types of defects in each subset without category imbalance or single scene problems. With a standardized preparation process, controllable defect generation conditions, unified preprocessing standards and high-precision manual annotations, the dataset can be widely applied in intelligent detection of water-based paint defects on wood products, industrial surface defect identification, and performance verification and optimization of lightweight object detection algorithms.

It provides high-quality, authentic and reliable data support for scientific research innovation and engineering application of automatic intelligent quality inspection technology for wood coating.