microsoft / microsoft/TRELLIS.2

[Question] Severe artifacts in TRELLIS.2 without background removal

Open
#65 0 comments 0 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

Hi @JeffreyXiang , thanks for your impressive work on TRELLIS.2 !

I noticed a significant quality degradation when testing TRELLIS.2 without background removal. The output exhibits severe artifacts and missing geometry (holes).

As a baseline, I tested the same input image (with background) using Hunyuan3D 2.1. It reconstructs the scene correctly without such issues. This suggests TRELLIS struggles to interpret the complex scene/background.

Image 1: TRELLIS.2 without background removal - Shows artifacts (the sofa)
Image 2: TRELLIS.2 with background removal - looks good
Image 3: Hunyuan3D 2.1 Result - Shows correct generation.

I'm wondering:

  1. Is this sensitivity caused by the o-voxel representation design or the training data distribution?
  2. Do you have any suggestions or workarounds to make TRELLIS robust to backgrounds (walls)?
Image Image

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

No source file, test, or entry point is identified. Start by reproducing the reported comparison between TRELLIS.2 with and without background removal using the same input image, then determine whether the issue can be narrowed to a documented model limitation or a reproducible defect.

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
20/100

Get new issues in your inbox

A short digest of beginner-friendly GitHub issues.