microsoft / microsoft/TRELLIS.2

Poor reconstruction of text / small details from image

Open
#57 0 comments 5 reactions 0 assignees View on GitHub

Nobody has claimed this yet.

Dominant language
Python
Stars
11.3k
Forks
1.4k
PR merge metrics
No merged PRs in 30d

Description

Overall results are good for organic shapes, but I’m consistently facing issues when the input image contains text, lettering, or fine details.

Image

The generated meshes tend to:

  • Distort letters heavily
  • Randomize strokes and edges
  • Lose sharp corners and inner cut-outs
  • Merge or deform small details beyond recognition

This happens even when the input image is clean, high-contrast, and well-lit.
Is this a known limitation of Trellis 2?

Contributor guide

No contributing guide indexed for this repository

First steps

  1. Read the whole issue, then the project's contributing guide.
  2. Comment on the issue to say you are picking it up — it saves two people doing the same work.
  3. Fork the repository and make your change on a branch.
  4. Open a pull request that references the issue number.

Research direction

The report does not identify a source file, model component, entry point, or test to investigate. Start by reproducing the behavior with the attached image and inspect the relevant TRELLIS.2 reconstruction pipeline. Done would require confirming whether text and fine-detail distortion is a known limitation or producing a scoped fix with regression coverage.

Written by the indexing model from the issue text.

Assessment

Tech stack
python
Domain
computer-graphics, machine-learning
Issue type
Bug
Difficulty
5/5
Estimated time
Over a week
Activity status
Stale
Clarity
Needs clarification
Newbie friendliness
25/100

Get new issues in your inbox

A short digest of beginner-friendly GitHub issues.