Welcome

This is SBXZY, trying to build a complete virtual world with visuals, cultures and actions.
The creation process is planned to be implemented mostly by artificial intelligence.

This might be wrong, but I believe that creating virtual worlds with human labor only is counterproductive.

 

Already implemented

• Parametricrized model generation (partially): There are Generative adversarial networks (GAN) methods that generate 3D models. But the neural network generator is limited by both effectiveness and efficiency. Instead, we use two primitives, graphs and pairs of edges, to generate all 3D models, so that artificial intelligence generating of models are easier.
  
• Procedual animations: Animations are just affine transformations if expressed in the Biovision hierarchical data (bvh) format. These transformations can be modeled by equations and parameters. This is implemented together with the parametricrized model generator.
  
• Interaction mechanism: Interaction mechanisms including physics simulations and visualization of virtual worlds are implemented by the Panda3D game engine.
  
• Music composer: Music composing is carried out in a 'semi automatic' manner. In GAN methods, users have little control of the generating process. In SBXZY's software, AI aids creating music sequences instead of doing it all, so that there is a higher chance AI can compose the type of music the user wants.
  
• Planet generator: This is done by many people many times, but we implemented with fractals and Perlin noises combined.

Planned

Purpose	Solution
Actions	Train parameters from existing captured animations.
Ecosystem	Evolutionary process to generate animals or plants.
Cultures	Evolve new cultures in new environments.

Examples

The parametricrized 3D model generator. Parametricrized but not yet automated.

Generate bvh animations from patameters. Generated animations can be used in Panda3D and Blender.

Planet terrains generator and visual explorer. Created an area of 5000km times 5000km with just 300MB of data.

A visual chatbot, demonstrating lip syncing for Panda3D, with expressions generated by GAN.

Related publications

[1] Xiang, Z. , Guo, Y. Controlling Melody Structures in Automatic Game Soundtrack Compositions With Adversarial Learning Guided Gaussian Mixture Models. IEEE transactions on Games (2021), 13,2 193-204. DOI: 10.1109/TG.2020.3035593
[2] Xiang, Z., Zhou, K., Guo, Y., Gaussian mixture noised random fractals with adversarial learning for automated creation of visual objects, Fractals - Complex geometry, patterns, and scaling in nature and society, (2020). 28,4. DOI: 10.1142/S0218348X20500681
[3] Xiang, Z. (2019). Towards automatic robot design: a black box function optimization approach. ChinaAvix. DOI: 10.12074/201905.00052
[4] Guo, Y., Fan, Y., Xiang, Z., Wang H., Meng, W., Xu, M. Zero-sample surface defect detection and classification based on semantic feedback neural network (2021) arXiv:2106.07959v1
[5] Xiang, Z. , Xiang C, Li T, Guo, Y. A self-adapting hierarchical actions and structures joint optimization framework for automatic design of robotic and animation skeletons. Softcomputing (2021), 25,3 DOI: 10.1007/s00500-020-05139-5.
[6] Xiang, Z., Xiao, Z., Wang, D., Xiao, J. (2017). Gaussian kernel smooth regression with topology learning neural networks and Python implementation. Neurocomputing, 260, 1-4. doi: http://dx.doi.org/10.1016/j.neucom.2017.01.051