MIO++

This dataset consists of images paired with corresponding segmentation labels, designed to enhance material understanding in 2D images. The images feature 12 semantic object categories, including furniture, vehicles, kitchenware, clothing, tools, and more. The segmentation labels include 30 distinct classes, providing fine-grained material categorization. These classes differentiate material types such as metal, wood, and plastic, further subdivided into surface properties like rough, smooth, and matte.This dataset is an invaluable resource for training deep learning models, offering rich prior knowledge to improve material perception and understanding in visual tasks.