(日本語版はこちら)
This is a repository for "character images search" from image and tags.
NOTE:
Added a development branch that supports image search on DuckDuckGo.
Please check HERE.
If you can use the conda command (Anaconda), you can install it with the command below.
conda create -n chara-searcher python=3.10
conda activate chara-searcher
conda install "pytorch<=2.1" torchvision torchaudio pytorch-cuda=11.8 -c pytorch -c nvidia
python install.py
After installation, you can run it with the following command.
conda activate chara-searcher
python standalone_ui.py
If you don't use conda, run python install.py
after installing Python and pytorch (and cuda).
How to install pytorch is HERE.
After installation, you can run it with python standalone_ui.py
.
Start webui and enter the following URL from "Install from URL" in "Extensions" to install.
https://github.com/NON906/chara-searcher.git
Click "Apply and restart UI" under "Extensions" -> "Installed", and restart.
Click "Upload Images Directory" (for image directories) or "Upload Video File" (for video files) and upload the file(s) to be searched.
After that, character extraction (background deletion), tagging, and vectorization for similar image searches will be automatically performed (please note that this may take some time).
Select the datas to search in "Target Datas".
You can search by specifying the following items.
After setting the following items, click on the search results to narrow down the search results.
Set the following items and click "Export" to export the displayed images.