The CC3D-PSE dataset (CAD Construction in 3D with Parametric Sharp Edges) is derived from the CC3D v1.0 (CAD Construction in 3D) dataset [1]. It contains 50k+ pairs of CAD models and their corresponding 3D scans. Each pair is also annotated with parametric sharp edges that were extracted from the CAD models. The parametric sharp edge annotations include three types of segments: circles, lines, and splines. The 3D objects in CC3D-PSE are unrestricted to any category, with the complexity varying from very basic to highly detailed models. The dataset can be implicitly split into  – (i) a subset of approximately 41K+ (~35K models for training ~6.3K models for testing) CAD models which has PSE as combinations of different entity types, and (ii) another subset of approximately 3K+ models (~2.4K models for training ~490 models for testing)  that has PSE as only straight line type.

Requesting CC3D-PSE

The CC3D-PSE dataset is available for use by external parties.  A license agreement must be requested and signed by the recipient and the research administration office director of your institution before the data can be provided. To make a request for the data, please contact us on Shapify3D (at) uni (dot) lu or use the following contact form.


(1) Once a license agreement is signed, we will give access to download the data.

(2) If this data is used, in whole or in part, the following paper must be referenced:


PVDeconv: Point-voxel deconvolution for autoencoding cad construction in 3d, Cherenkova, Kseniya, Djamila Aouada, and Gleb Gusev, 2020 IEEE International Conference on Image Processing (ICIP).

SEPIC-Net (Details Coming Soon!)