3dlg-hcvc/vigil3d

ViGiL3D数据集是一个用于3D视觉定位的多样化语言数据集。它包含了ScanNet和ScanNet++的场景提示，并且每个描述都包含目标对象和锚点对象的属性和关系信息，如颜色、大小、形状、材料、纹理、功能、风格、文本标签、状态等属性，以及对象之间的远近、方向、包含、排列、序数、比较等关系。此外，数据集还包含了关于描述语言模式的元数据属性，如描述中的属性和关系数量、名词的特异性等。

The ViGiL3D dataset is a linguistically diverse dataset for 3D visual grounding. It includes scene prompts for both ScanNet and ScanNet++, and each description contains information about the attributes and relationships of the target and anchor objects, such as color, size, shape, material, texture, function, style, text label, state, and relationships like nearness, direction, containment, arrangement, ordinal position, comparison, etc. In addition, the dataset includes metadata attributes about the language patterns in the descriptions, such as the number of attributes and relationships, the specificity of the nouns used, etc.