mirror of
https://github.com/kohya-ss/sd-scripts.git
synced 2026-04-06 13:47:06 +00:00
docs: add skip_image_resolution to config README (#2288)
* docs: add skip_image_resolution option to config README Document the skip_image_resolution dataset option added in PR #2273. Add option description, multi-resolution dataset TOML example, and command-line argument entry to both Japanese and English config READMEs. Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com> * docs: clarify `skip_image_resolution` functionality in dataset config --------- Co-authored-by: Claude Opus 4.6 <noreply@anthropic.com>
This commit is contained in:
Kohya S.
GitHub
@@ -122,11 +122,15 @@ These are options related to the configuration of the data set. They cannot be d
|
||||
| `max_bucket_reso` | `1024` | o | o |
|
||||
| `min_bucket_reso` | `128` | o | o |
|
||||
| `resolution` | `256`, `[512, 512]` | o | o |
|
||||
| `skip_image_resolution` | `768`, `[512, 768]` | o | o |
|
||||
|
||||
* `batch_size`
|
||||
* This corresponds to the command-line argument `--train_batch_size`.
|
||||
* `max_bucket_reso`, `min_bucket_reso`
|
||||
* Specify the maximum and minimum resolutions of the bucket. It must be divisible by `bucket_reso_steps`.
|
||||
* `skip_image_resolution`
|
||||
* Images whose original resolution (area) is equal to or smaller than the specified resolution will be skipped. Specify as `'size'` or `[width, height]`. This corresponds to the command-line argument `--skip_image_resolution`.
|
||||
* Useful when sharing the same image directory across multiple datasets with different resolutions, to exclude low-resolution source images from higher-resolution datasets.
|
||||
|
||||
These settings are fixed per dataset. That means that subsets belonging to the same dataset will share these settings. For example, if you want to prepare datasets with different resolutions, you can define them as separate datasets as shown in the example above, and set different resolutions for each.
|
||||
|
||||
@@ -254,6 +258,34 @@ resolution = 768
|
||||
image_dir = 'C:\hoge'
|
||||
```
|
||||
|
||||
When using multi-resolution datasets, you can use `skip_image_resolution` to exclude images whose original size is too small for higher-resolution datasets. This prevents overlapping of low-resolution images across datasets and improves training quality. This option can also be used to simply exclude low-resolution source images from datasets.
|
||||
|
||||
```toml
|
||||
[general]
|
||||
enable_bucket = true
|
||||
bucket_no_upscale = true
|
||||
max_bucket_reso = 1536
|
||||
|
||||
[[datasets]]
|
||||
resolution = 768
|
||||
[[datasets.subsets]]
|
||||
image_dir = 'C:\hoge'
|
||||
|
||||
[[datasets]]
|
||||
resolution = 1024
|
||||
skip_image_resolution = 768
|
||||
[[datasets.subsets]]
|
||||
image_dir = 'C:\hoge'
|
||||
|
||||
[[datasets]]
|
||||
resolution = 1280
|
||||
skip_image_resolution = 1024
|
||||
[[datasets.subsets]]
|
||||
image_dir = 'C:\hoge'
|
||||
```
|
||||
|
||||
In this example, the 1024-resolution dataset skips images whose original size is 768x768 or smaller, and the 1280-resolution dataset skips images whose original size is 1024x1024 or smaller.
|
||||
|
||||
## Command Line Argument and Configuration File
|
||||
|
||||
There are options in the configuration file that have overlapping roles with command line argument options.
|
||||
@@ -284,6 +316,7 @@ For the command line options listed below, if an option is specified in both the
|
||||
| `--random_crop` | |
|
||||
| `--resolution` | |
|
||||
| `--shuffle_caption` | |
|
||||
| `--skip_image_resolution` | |
|
||||
| `--train_batch_size` | `batch_size` |
|
||||
|
||||
## Error Guide
|
||||
|
||||
@@ -115,11 +115,15 @@ DreamBooth の手法と fine tuning の手法の両方とも利用可能な学
|
||||
| `max_bucket_reso` | `1024` | o | o |
|
||||
| `min_bucket_reso` | `128` | o | o |
|
||||
| `resolution` | `256`, `[512, 512]` | o | o |
|
||||
| `skip_image_resolution` | `768`, `[512, 768]` | o | o |
|
||||
|
||||
* `batch_size`
|
||||
* コマンドライン引数の `--train_batch_size` と同等です。
|
||||
* `max_bucket_reso`, `min_bucket_reso`
|
||||
* bucketの最大、最小解像度を指定します。`bucket_reso_steps` で割り切れる必要があります。
|
||||
* `skip_image_resolution`
|
||||
* 指定した解像度（面積）以下の画像をスキップします。`'サイズ'` または `[幅, 高さ]` で指定します。コマンドライン引数の `--skip_image_resolution` と同等です。
|
||||
* 同じ画像ディレクトリを異なる解像度の複数のデータセットで使い回す場合に、低解像度の元画像を高解像度のデータセットから除外するために使用します。
|
||||
|
||||
これらの設定はデータセットごとに固定です。
|
||||
つまり、データセットに所属するサブセットはこれらの設定を共有することになります。
|
||||
@@ -259,6 +263,34 @@ resolution = 768
|
||||
image_dir = 'C:\hoge'
|
||||
```
|
||||
|
||||
なお、マルチ解像度データセットでは `skip_image_resolution` を使用して、元の画像サイズが小さい画像を高解像度データセットから除外できます。これにより、低解像度画像のデータセット間での重複を防ぎ、学習品質を向上させることができます。また、小さい画像を除外するフィルターとしても機能します。
|
||||
|
||||
```toml
|
||||
[general]
|
||||
enable_bucket = true
|
||||
bucket_no_upscale = true
|
||||
max_bucket_reso = 1536
|
||||
|
||||
[[datasets]]
|
||||
resolution = 768
|
||||
[[datasets.subsets]]
|
||||
image_dir = 'C:\hoge'
|
||||
|
||||
[[datasets]]
|
||||
resolution = 1024
|
||||
skip_image_resolution = 768
|
||||
[[datasets.subsets]]
|
||||
image_dir = 'C:\hoge'
|
||||
|
||||
[[datasets]]
|
||||
resolution = 1280
|
||||
skip_image_resolution = 1024
|
||||
[[datasets.subsets]]
|
||||
image_dir = 'C:\hoge'
|
||||
```
|
||||
|
||||
この例では、1024 解像度のデータセットでは元の画像サイズが 768x768 以下の画像がスキップされ、1280 解像度のデータセットでは 1024x1024 以下の画像がスキップされます。
|
||||
|
||||
## コマンドライン引数との併用
|
||||
|
||||
設定ファイルのオプションの中には、コマンドライン引数のオプションと役割が重複しているものがあります。
|
||||
@@ -289,6 +321,7 @@ resolution = 768
|
||||
| `--random_crop` | |
|
||||
| `--resolution` | |
|
||||
| `--shuffle_caption` | |
|
||||
| `--skip_image_resolution` | |
|
||||
| `--train_batch_size` | `batch_size` |
|
||||
|
||||
## エラーの手引き
|
||||
|
||||
Reference in New Issue
Block a user