PQ-NET: A Generative Part Seq2Seq Network for 3D Shapes
This work addresses the challenge of generating structured 3D shapes for applications in computer graphics and robotics, but it appears incremental as it builds on existing Seq2Seq and autoencoder methods.
The authors tackled the problem of generating 3D shapes by introducing PQ-NET, a deep neural network that represents and generates shapes through sequential part assembly, enabling tasks like autoencoding, interpolation, novel shape generation, and single-view 3D reconstruction with meaningful parts.
We introduce PQ-NET, a deep neural network which represents and generates 3D shapes via sequential part assembly. The input to our network is a 3D shape segmented into parts, where each part is first encoded into a feature representation using a part autoencoder. The core component of PQ-NET is a sequence-to-sequence or Seq2Seq autoencoder which encodes a sequence of part features into a latent vector of fixed size, and the decoder reconstructs the 3D shape, one part at a time, resulting in a sequential assembly. The latent space formed by the Seq2Seq encoder encodes both part structure and fine part geometry. The decoder can be adapted to perform several generative tasks including shape autoencoding, interpolation, novel shape generation, and single-view 3D reconstruction, where the generated shapes are all composed of meaningful parts.