Point Cloud Generation with Continuous Conditioning


Larissa T. Triess1,2 Andre Bühler1,3 David Peter1 Fabian B. Flohr1,3 J. Marius Zöllner2,4

1Mercedes-Benz AG 2Karlsruhe Institute
of Technology 		3University of Stuttgart 4Research Center for
Information Technology

In 2022 International Conference on Artificial Intelligence and Statistics (AISTATS)

[Paper]


metric
Proposed Approach:
Our method can generate a diverse set of object shapes and be conditioned on object dimensions (values on the axes). The figure shows objects generated from the same latent vector z, but with different continuous conditioning parameters y. The generated objects are realistic and semantically meaningful. Our method generalizes to out-of-distribution dimensions (outside of dotted shape).


Abstract

Generative models can be used to synthesize 3D objects of high quality and diversity. However, there is typically no control over the properties of the generated object. This paper proposes a novel GAN setup that generates 3D point cloud shapes conditioned on a continuous parameter. In an exemplary application, we use this to guide the generative process to create a 3D object with a custom-fit shape. We formulate this generation process in a multi-task setting by using the concept of auxiliary classifier GANs. Further, we propose to sample the generator label input for training from a KDE of the dataset. Our ablations show that this leads to significant performance increase in regions with few samples. Extensive quantitative and qualitative experiments show that we gain explicit control over the object dimensions while maintaining good generation quality and diversity.


Paper





Paper: [ArXiv]

Citation: 
@inproceedings{triess2022aistats,
  title = {
	Point Cloud Generation with Continuous Conditioning
  },
  author = {
    Larissa T. Triess and
    Andre B\"uhler and
    David Peter and
    Fabian B. Flohr and
    J. Marius Z\"ollner
  },
  booktitle = {
    Proc. of the International Conference on Artificial Intelligence and Statistics (AISTATS)
},
  year = {2022},
}


Contributions

  • We propose a GAN setup which formulates the generation process conditioned on continuous parameters in a multi-task setting by adapting the concept of auxiliary classifier GANs (Odena et al.,2017).
  • We propose to sample the generator conditioning input for training from a KDE of the parameter distribution. Our ablations show that this leads to a significant performance increase in regions with few samples.
  • We provide a number of qualitative evaluations that show that we gain explicit control over the object dimensions while maintaining generation quality and diversity.


Qualitative Results

chair chair chair chair chair chair
table table table table table table
airplane airplane lamp lamp car car


References

Odena, A., Olah, C., an Shlens, J. (2017). Conditional Image Synthesis with Auxiliary Classifier GANs. In Proceedings of the International Conference on Machine Learning.