kohya-ss / kohya-ss/sd-scripts

Selection of scaling resampler

Open
#167 1 comment 0 reactions 0 assignees View on GitHub
Dominant language
Python
Stars
7.2k
Forks
1.2k
Avg merge
11m
Merged PRs (30d)
2

Description

Thank you for your hard work.

It would be nice to add a function to select a resampler for Scaling, or replace the default resampler.

Resampler is currently set to INTER_AREA.

In general, the LANCZOS family for upscale and the BICUBIC family for downscale are known to be the best in terms of speed and efficiency to maintain quality. (based on Adobe's products)

Of course, it's right to do this in the process of preparing the dataset, but I think there are times when we have no choice but to use resampler in the process of converting to latent.

If possible, I would like to make these samplers available.

Contributor guide

No contributing guide indexed for this repository

Research direction

Locate the scaling path that currently uses INTER_AREA and read how images are converted to latent. Determine where a selectable or replaceable resampler can be exposed, including LANCZOS and BICUBIC, then verify that scaling still works for dataset and latent-conversion workflows.

Written by the indexing model from the issue text.

Assessment

Tech stack
python
Domain
machine-learning
Issue type
Feature
Difficulty
4/5
Estimated time
3-5 days
Activity status
Stale
Clarity
Mostly clear
Newbie friendliness
35/100

Get new issues in your inbox

A short digest of beginner-friendly GitHub issues.