Machine learning techniques for sketch-to-3D shape generation

One or more non-transitory computer-readable media storing instructions that, when executed by one or more processors, cause the one or more processors to perform the steps of:

• generating semantic features associated with an input sketch;
• generating, using a generative machine learning model, a plurality of predicted shape embeddings from a set of fully masked shape embeddings based on the semantic features associated with the input sketch;
• and
• converting the predicted shape embeddings into one or more 3D shapes.

A computer-implemented method for training a machine learning model to perform three-dimensional (3D) shape generation, the computer-implemented method comprising:

• generating a set of shape embeddings associated with a 3D object;
• generating semantic features associated with a two-dimensional (2D) rendering of the 3D object;
• generating a set of masked shape embeddings by masking a portion of the set of shape embeddings;
• generating, using a generative machine learning model, a plurality of predicted shape embeddings based on the masked shape embeddings and the textual semantic features associated with the 2D rendering;
• and
• calculating a masking loss function based on the predicted shape embeddings, the set of shape embeddings associated with the 3D object, and the semantic features.