When a folder is specified in "Input path", the converted file is saved in the specified folder (without changing the folder structure). If the specified folder does not exist, it will be created automatically.
For magnifications greater than 2x (only one time if removing noise is done the first time) Enlarges by 2x until the specified magnification is exceeded, and if the magnification is not a power of 2, shrinks last The process of doing is performed. Therefore, the conversion result may be a flat picture.
* 2D illustration (RGB model): 2D illustration model that converts all RGB of the image
* Photo/Anime (Photo model): Model for photo/animation
* 2D illustration (UpRGB model): A model that converts faster than 2D illustration (RGB model) with the same or higher image quality. However, the amount of memory (VRAM) that is consumed is larger than that of the RGB model, so adjust the division size if forced termination during conversion.
* Photo/animation (Up Photo model): A model that converts at higher speed than the photo/animation (Photo model) with the same or higher image quality. However, the amount of memory (VRAM) that is consumed is larger than that of the Photo model, so adjust the division size if it is forcibly terminated during conversion.
* 2D illustration (Y model): A model for 2D illustration that converts only the brightness of the image
* 2D illustration (UpResNet10 model): A model that converts with higher image quality than 2D illustration (UpRGB model). Note that this model will change the output result if the division size is different.
* 2D illustration (CUnet model): The model that can convert the 2D illustration with the highest image quality with the included model. Note that this model will change the output result if the division size is different.
Specify the width (pixel unit) when dividing and processing internally.
How to decide the optimum number (the process ends at the fastest) is explained in the section "Split size".
Separated by "-------", the upper one is the divisor of the vertical and horizontal size of the input image,
The lower one is a general division size read from "crop_size_list.txt".
If the partition size is too large, the amount of memory required (VRAM amount when using GPU) exceeds the memory available on the PC, and the software will be killed, so be careful.
Since it will affect the processing speed to some extent, when converting a large number of images with the same image size by specifying a folder, it is recommended to check the optimal division size before conversion.
However, be aware that depending on the model, the output result may change when the division size is changed.
(In that case, use the default split size and adjust the batch size to speed up processing.)
### Batch size
Specify the size when processing all at once internally.
Increasing the batch size may increase the processing speed.
Make sure that the amount of memory required, as well as the partition size, does not exceed the memory available on your PC.
Set whether to start the conversion automatically when the input file is specified by the reference button or drag and drop.
If the input file is given to exe as an argument, the setting contents of this item have no effect.
* Do not start automatically: Do not start conversion automatically when inputting a file
* Start after inputting one file: Start conversion automatically after inputting one file
* Start after inputting a folder or multiple files: Start conversion automatically when inputting a folder or multiple files. Single image file Convert files only when adjusting conversion settings
However, there is a time lag before it actually stops.
The progress bar shows the progress when changing multiple images.
The log shows the estimated remaining time, which is an estimate when processing multiple files with the same height and width.
Therefore, it is not useful when the size of the file is different, and when the number of images to be processed is 2 or less, only "Unknown" is displayed.
Specify the mini-batch size. The default value is `1`.
The mini-batch size is the number of blocks that the image is divided into by "division size" and processed at the same time. For example, if you specify `2`, it will be converted every 2 blocks.
When the mini-batch size is increased, the GPU usage rate increases as well as when the split size is increased, but if you feel it is measured, it is more effective to increase the split size.
(For example, if the split size is `64` and the mini-batch size is `4`, the split size is `128` and the mini-batch size is `1`.
The image is divided into pieces of a certain size, converted one by one, and finally combined into a single image.
The division size (crop_size) is the width (in pixels) when dividing this image.
If the GPU is not exhausted even if it is converted with CUDA (the usage of the GPU is not close to 100%),
If you increase this number, the process may end sooner. (Because the GPU can be used up)
Please adjust it while seeing GPU Load (GPU usage rate) and Memory Used (VRAM usage rate) with [GPU-Z](http://www.techpowerup.com/gpuz/).
Also, refer to the following characteristics.
* Larger numbers do not necessarily mean faster
* If the division size is a divisor (or a number with a small remainder when divided) of the vertical and horizontal size of the image, the amount of wasteful calculation decreases and the speed increases. (In some cases, it seems that the numerical value that does not apply to this condition is the fastest.)
* If you double the number, theoretically the amount of memory used will be 4 times (actually it is 3 to 4 times), so be careful not to drop the software. Especially, CUDA consumes much more memory than cuDNN, so be careful.
About images with alpha channel
--------
This software also supports enlargement of images with alpha channel.
However, please note that it takes about twice as long as the image without alpha channel is enlarged because the process is to enlarge the alpha channel by itself.
However, if the alpha channel is composed of a single color, it can be expanded in about the same time as it was without.
The format of language files
--------
Language files format is JSON.
If you create new language file, add language setting to'lang/LangList.txt'.
'lang/LangList.txt' format is TSV(Tab-Separated Values).
* LangName: Language name
* LangID: Primary language [See MSDN](https://msdn.microsoft.com/en-us/library/windows/desktop/dd318693.aspx)
The original [waifu2x](https://github.com/nagadomi/waifu2x) and model were produced and published under the MIT license [ultraist](https://twitter.com/ultraistter) Mr.
Created [waifu2x-converter](https://github.com/WL-Amigo/waifu2x-converter-cpp) based on the original waifu2x [Amigo](https://twitter.com/WL_Amigo) Mr. (I used to refer to how to write README and LICENSE.txt, how to use OpenCV)
[waifu2x-chainer](https://github.com/tsurumeso/waifu2x-chainer) was created to create the original model, and it was published under the MIT license [tsurumeso](https:// github.com/tsurumeso)
will be grateful to.
Also, @paul70078 for translating the message into English, @yoonhakcher for translating the message into Simplified Chinese, @mzhboy for the pull request for Simplified Chinese translation,
@Kenin0726 for translating the message into Korean, @aruhirin for suggesting improvements to the Korean translation,
@Lizardon1995 for translating messages in Traditional Chinese, @yoonhakcher, @Scharynche for pull request for Turkish translation, @Serized for pull request for French translation, Brazilian Portuguese @Simrafael for pull request for translation, @AndreasWebdev for pull request for German translation, @07pepa for pull request for Czech translation,
Thanks to JYUNYA for providing the GUI version of the icon.
