Materials In Paintings (MIP)

Materials In Paintings (MIP)数据集是由荷兰代尔夫特理工大学的感知智能实验室和美国康奈尔大学的计算机科学系共同创建的，旨在研究绘画中材料的描绘与感知。该数据集包含19,325幅绘画，通过众包方式标注了超过20万个边界框，并自动提取了多边形段。每个边界框都被赋予了一个粗略标签和一个细粒度标签，如'织物'和'天鹅绒'。MIP数据集不仅适用于艺术史研究，还能为人类感知和计算机视觉领域的研究提供丰富的资源，帮助解决材料描绘和感知的跨学科问题。

Materials In Paintings (MIP) Dataset was co-developed by the Perceptual Intelligence Lab at Delft University of Technology in the Netherlands and the Department of Computer Science at Cornell University in the United States, aiming to investigate the depiction and perception of materials in paintings. This dataset includes 19,325 paintings, with over 200,000 bounding boxes annotated via crowdsourcing and polygonal segments automatically extracted. Each bounding box is assigned both a coarse-grained label and a fine-grained label, such as "fabric" and "velvet". The MIP dataset is not only suitable for art historical research, but also offers a rich resource for studies in the fields of human perception and computer vision, assisting in addressing interdisciplinary questions concerning material depiction and perception.