Proposed: Modify the Mesh Transformer for Quads

The Context

We’re working on a project that involves using an autoencoder/transformer for 3D models. The current setup does not have a setting for quads, which is a requirement for our project.

The Problem Statement

The challenge is to modify the autoencoder/transformer to include a setting for quads and achieve a loss around 0.35.

Describe how your proposal will work with code, pseudo-code, mock-ups, or diagrams

  1. Modify the Autoencoder/Transformer: Update the existing code to include a quads=True parameter in the autoencoder/transformer.

  2. Test the Modification: Run tests to ensure that the modification works as expected and doesn’t break existing functionality.

  3. Optimize the Loss: Adjust the parameters and settings of the autoencoder/transformer to aim for a loss around 0.35 when processing quads.

The Benefits

  • This modification will allow us to process quads in our 3D models.
  • Achieving a loss around 0.35 will improve the quality of our model transformations.

The Downsides

  • Modifying the existing code could introduce new bugs or issues.
  • It might be challenging to achieve the desired loss.

The Road Not Taken

An alternative approach could be to convert quads into triangles before processing them with the autoencoder/transformer. However, this might result in loss of information or quality.

The Infrequent Use Case

This modification might not be necessary if we are dealing with 3D models that do not contain quads.

In Core and Done by Us

The V-Sekai development team will handle the modification of the autoencoder/transformer and the optimization of the loss.

Status

Status: Proposed

Decision Makers

  • V-Sekai development team

Tags

  • V-Sekai

Further Reading

  1. V-Sekai · GitHub - Official GitHub account for the V-Sekai development community focusing on social VR functionality for the Godot Engine.
  2. V-Sekai/v-sekai-game - The GitHub page for the V-Sekai open-source project, which brings social VR/VRSNS/metaverse components to the Godot Engine.

AI assistant Aria assisted with this proposal.