nagadomi
commented
5 years ago | model_dir | Resource ID | English Label |
|---|---|---|
| models/upconv_7_anime_style_art_rgb | IDC_RADIO_MODEL_UPCONV_RGB | 2-D illust (UpRGB Model) |
| models/upconv_7_photo | IDC_RADIO_MODEL_UPCONV_PHOTO | Photo (UpPhoto Model) |
| models/anime_style_art_rgb | IDC_RADIO_MODEL_RGB | 2-D illust (RGB Model) |
| models/photo | IDC_RADIO_MODEL_PHOTO | Photography, Anime |
| models/anime_style_art | IDC_RADIO_MODEL_Y | 2-D illust (Y Model) |
| models/upresnet10 | IDC_RADIO_MODEL_UpResNet10 | 2-D illust (UpResNet10) |
| models/cunet | IDC_RADIO_MODEL_CUNET | 2-D illust (CUnet Model) |
From code. https://github.com/lltcggie/waifu2x-caffe/blob/8b53d0f086f44723cb6e101c12e42e63fb4c6a63/waifu2x-caffe-gui/MainDialog.h#L39-L69 https://github.com/lltcggie/waifu2x-caffe/blob/master/bin/lang/english.json
model/ukbench is prototype of photo model made long time ago, just ignore it.
Joachim-Otahal
Command line model models\upconv_7_anime_style_art_rgb gives me the best result for my current batch. But which GUI-Model is upconv_7_anime_style_art_rgb? I tried them all seven of them and the result looked different. I also notice I have eight models on the cmdline version, so that exact model might be missing in the GUI version... Apart from that: Running very stable and very well here, > 200'000 frames already upsampled...