Dense Material Segmentation dataset (DMS)
收藏资源简介:
Dense Material Segmentation dataset (DMS) 是由苹果公司创建的一个大规模室内外场景解析材料分割数据集。该数据集包含44,560张图像,共计320万个密集标注的材料段,覆盖了更广泛的场景、物体、视角和材料类型。DMS数据集旨在通过提供丰富的材料标注,推动材料识别技术的发展,特别是在室内外场景解析领域。数据集的创建过程涉及详细的图像选择、人工标注和多轮验证,确保了标注的准确性和一致性。DMS数据集的应用领域广泛,包括机器人导航、材料分配、触觉模拟和上下文感知混合现实等,旨在解决现有模型在特定环境下性能不足的问题。
The Dense Material Segmentation dataset (DMS) is a large-scale dataset for material segmentation and scene parsing of indoor and outdoor scenes, developed by Apple Inc. It contains 44,560 images with a total of 3.2 million densely annotated material segments, covering a broad spectrum of scenes, objects, viewpoints, and material categories. The DMS dataset aims to promote the advancement of material recognition technologies, especially in the domain of indoor and outdoor scene parsing, by providing rich material annotations. The development process of the dataset entails meticulous image selection, manual annotation, and multi-round validation, ensuring the accuracy and consistency of the annotations. The DMS dataset has a wide range of application scenarios, including robot navigation, material allocation, haptic simulation, and context-aware mixed reality, among others, and is designed to address the performance shortcomings of existing models in specific environments.




