search

deepghs / waifuc

Efficient Train Data Collector for Anime Waifu
https://deepghs.github.io/waifuc/
MIT License
265 stars 20 forks source link

Able to choose to filter certain file formats, such as videos/GIFs/images #56

Open wochenlong opened 5 months ago

wochenlong commented 5 months ago

When processing resources from a website or locally, I would like to be able to choose to filter certain file formats, such as videos/GIFs/images. when using waifuc to retrieve images from Danbooru, is it possible to filter out resources other than images? Extracting usable images from MP4 files consumes a lot of my CPU time—but in reality, I don't need them. We all know that Danbooru supports JPEG, PNG (including Animated PNG), GIF, SWF, and WEBM. However, sometimes when there are a sufficient number of resources for a particular character, only fetching images can save on the speed of resource acquisition. ece902d8a348a7ed2eb10144aca8448b

wochenlong commented 5 months ago

There is another scenario where, for example, when I want to train the PV style of "blue_archive" I would like to only fetch and process video file formats. In this case, I can exclude file formats other than videos.