Heater Tank Inner Wall dataset

The HTIW (Hot-water Tank Inner Wall) dataset is a novel collection of images capturing the inner enamel surfaces of water heater tanks. All 360 images were acquired on a live industrial production line using our specialized IAR system. To ensure representative data across various tank models and sizes, collection parameters and lighting angles were manually optimized under the guidance of professional quality inspectors.The dataset specifically focuses on the challenge of non-uniform low-light conditions. Through quantitative analysis using Principal Component Analysis (PCA) on combined grayscale and brightness features, we identified three distinct clusters in the data. Consequently, the dataset was classified via K-means clustering into three representative non-uniform lighting patterns:Uniform Low-Light (primarily tank top\/bottom)Center-Bright, Edge-Dark (primarily tank sidewalls)Center-Dark, Edge-Bright (primarily weld seam areas)The final dataset contains 360 high-resolution (1440\u00d71080) images, which have been carefully selected to exclude high-noise or severely blurred samples while preserving the authentic characteristics of the raw captures. The dataset is divided into a training set of 288 images and a test set of 72 images, maintaining an 8:2 ratio within each of the three lighting categories to mirror real-world distribution.To our knowledge, HTIW is the first publicly available image dataset of water heater tank interiors. Its key advantages are:Uniqueness: It provides comprehensive coverage, including the top, bottom, and sidewalls, filling a critical gap left by existing datasets.Authenticity: The images are captured from a real-world industrial environment without any pre-processing, ensuring they accurately reflect the complex lighting challenges and true physical surface conditions.Challenge: The diversity of severe non-uniform illumination makes it a challenging and valuable benchmark for developing and testing image enhancement algorithms.