Ecosyste.ms: Awesome
An open API service indexing awesome lists of open source software.
https://github.com/usu171/pximse
Local anime style image search based on DeepDanbooru, CLIP and PaddleOCR
https://github.com/usu171/pximse
anime clip deepdanbooru nuxtjs python
Last synced: 25 days ago
JSON representation
Local anime style image search based on DeepDanbooru, CLIP and PaddleOCR
- Host: GitHub
- URL: https://github.com/usu171/pximse
- Owner: Usu171
- License: mit
- Created: 2024-11-25T09:04:32.000Z (30 days ago)
- Default Branch: master
- Last Pushed: 2024-11-26T06:58:07.000Z (29 days ago)
- Last Synced: 2024-11-26T07:32:32.769Z (29 days ago)
- Topics: anime, clip, deepdanbooru, nuxtjs, python
- Language: Python
- Homepage: https://usu171.uk/2024/11/25/PxIMSE/
- Size: 321 KB
- Stars: 2
- Watchers: 2
- Forks: 0
- Open Issues: 0
-
Metadata Files:
- Readme: README.md
- License: LICENSE
Awesome Lists containing this project
README
# PxIMSE
Local anime style image search based on [DeepDanbooru](https://github.com/KichangKim/DeepDanbooru), [CLIP](https://github.com/openai/CLIP) and [PaddleOCR](https://paddlepaddle.github.io/PaddleOCR/main/index.html)
### Supported file name formats
Images from [Pixiv](https://www.pixiv.net/), [PixivBatchDownloader](https://github.com/xuejianxianzun/PixivBatchDownloader), [PixEZ](https://github.com/Notsfsssf/pixez-flutter) ... , [Twitter Media Downloader](https://greasyfork.org/zh-CN/scripts/423001-twitter-media-downloader), [yande.re](https://yande.re/) or other sources, but without ID information:
- pixiv: {user_id}_{id}_p{index} **(Recommended)**
- pixiv: {id}_p{index}
- twitter: twitter_{user_name}(@{user_id})\_{date}\_{id}_photo
- yande.re: yande.re {id} {tags}
## Pre-requisites
- [Python 3.10+](https://www.python.org/downloads/)
- [Node.js 20+](https://nodejs.org/en/)
- [Nginx](https://nginx.org/)
- [Pre-trained DeepDanbooru model](https://github.com/KichangKim/DeepDanbooru/releases/tag/v3-20211112-sgd-e28)
### Databases
- [MongoDB](https://www.mongodb.com/)
- [Milvus](https://milvus.io/)
Refer to `docker-compose.yml`
### Dependencies
```sh
pip3 install -r requirements.txt
```
You might need to handle torch, tensorflow, and paddlepaddle manually
```sh
pnpm install
```
### configuration
Configure `python/config.yml` and `config.mjs`
## Usage
### 1. Import image information
Import image information, DeepDanbooru and OCR results to MongoDB
```sh
python import_images.py mongo
```
Import CLIP results to Milvus
```sh
python import_images.py milvus
```
Get information from Pixiv website (optional)
```sh
python import_images.py pixiv
```
Attempt to assign the username from an existing document in the database to a document with the same user ID that doesn't have a username (Indexes must be created on the username and userid fields) (optional)
```sh
python import_images.py update
```
### 2. Start Service
```sh
cd python
uvicorn api:app --host 0.0.0.0 --port 8000
```
```sh
pnpm dev
```
### Format of Query Criteria
**Exists:**
tags.xxx
tags1:xxx
**Does not exist:**
!tags.xxx
!tags1:xxx
**Greater than, less than, equal to, not equal to:**
filesize;>1000000
tags.xxx;<=0.9
date;>2000-01-01&<2010-01-01
likeCount;=100
viewCount;!=200
**Text length:**
textlen;>xx
tags are obtained from DeepDanbooru
tags1 are retrieved from Pixiv website (For yandere, from the file name)
## Credits
Inspired by:
- [clip-image-search](https://github.com/atarss/clip-image-search)