microsoft / microsoft/TRELLIS

preservation of details of input image

Open
#231 1 comment 0 reactions 0 assignees View on GitHub

Nobody has claimed this yet.

Dominant language
Python
Stars
13.7k
Forks
1.3k
PR merge metrics
No merged PRs in 30d

Description

Hi, I have tried some generation-based methods to get 3d model but found that many of them including trellis could not preserve the details of condition image, for example, sharp edge was losing. With my own understanding, input image having strong condition, should guide model to generate similiar result with input.

Is one of module of the pipeline that drops the infomation? Is any advice to improve preserve the details when finetuning model?

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 issue names no files, tests, or entry points. Start by examining the TRELLIS image-to-3D pipeline and the conditioning path, then investigate where sharp input-image details are lost during generation or fine-tuning; done would require identifying the responsible stage and demonstrating improved detail preservation.

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.