microsoft / microsoft/TRELLIS

How many samples for fine-tuning?

Open
#261 0 comments 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 everyone,

I am trying to use TRELLIS on a specific type of object and want to generate accurate 3D models from multiple input images. For some objects, I don't get proper outcomes with different configs and seeds. I wonder how many samples I can use to improve output quality with fine-tuning. I can generate around 1k samples using SolidWorks. Is it enough? My objects are kinds of pipes with different curvatures that make it hard for the model to reconstruct an accurate 3D output.

Also, it would be great if you have a good configuration (seed, steps, etc.) to share here.

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 files, tests, or entry points are mentioned. Start by reviewing the repository's fine-tuning guidance and the reported SolidWorks dataset and pipe-shape problem; the issue is complete when it has an evidence-based recommendation for sample count and configuration settings such as seed and steps.

Written by the indexing model from the issue text.

Assessment

Tech stack
machine-learning, python
Domain
computer-graphics, machine-learning
Issue type
Feature
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.