fixes typo in tagging methodology doc
Browse files
kazusa/tagging methodology.md
CHANGED
|
@@ -17,7 +17,7 @@ Yes, it takes forever to do all of this shit. No, I don't recommend it unless yo
|
|
| 17 |
|
| 18 |
- Scraped `1girl kazusa_(blue_archive) order:popularity` from sancom, curated for quality, then exported from Hydrus to feed into WD1.4 Tagger.
|
| 19 |
- Kazusa has a shitload of good art so I had to be very picky to get down to 280 images, which is still a lot. In hindsight I think huge datasets aren't really a problem; they let you train for longer without overfitting.
|
| 20 |
-
- Gelbooru is probably fine too. Danbooru sucks for
|
| 21 |
- I also got a few newer images from pixiv, don't remember which ones.
|
| 22 |
- Exported final images from Hydrus to feed into WD1.4 Tagger
|
| 23 |
- Auto-tagged with WD1.4 Swinv2 at 0.25 confidence
|
|
|
|
| 17 |
|
| 18 |
- Scraped `1girl kazusa_(blue_archive) order:popularity` from sancom, curated for quality, then exported from Hydrus to feed into WD1.4 Tagger.
|
| 19 |
- Kazusa has a shitload of good art so I had to be very picky to get down to 280 images, which is still a lot. In hindsight I think huge datasets aren't really a problem; they let you train for longer without overfitting.
|
| 20 |
+
- Gelbooru is probably fine too. Danbooru sucks for ロリ unless you have Gold.
|
| 21 |
- I also got a few newer images from pixiv, don't remember which ones.
|
| 22 |
- Exported final images from Hydrus to feed into WD1.4 Tagger
|
| 23 |
- Auto-tagged with WD1.4 Swinv2 at 0.25 confidence
|