Welcome to the 3DCoMPaT++ documentation. 3DCoMPaT++ is a richly annotated large-scale dataset of 10.000 unique models, with a thousand generated styles per model.
Models are annotated at the part-instance level following a consistent nomenclature hand-defined for each of the 42 shape categories, at two different semantic levels: coarse-grained and fine-grained.
Each object with the applied part-material compositions is rendered from four equally spaced views as well as four randomized views. For each view, we also provide depth maps, part segmentation masks, material segmentation masks and rendering camera parameters.
For any question regarding 3DCoMPaT++, please open a ticket in our GitHub repository:
To read more about our dataset, navigate our table of contents: