https://github.com/kuzulabz/waifutagger

A utility app that tags your images with danbooru tags - fully offline
https://github.com/kuzulabz/waifutagger

ai android-app image-classification onnxruntime react-native weeb

Last synced: about 1 month ago
JSON representation

A utility app that tags your images with danbooru tags - fully offline

• Host: GitHub
• URL: https://github.com/kuzulabz/waifutagger
• Owner: KuzuLabz
• Created: 2024-03-07T01:42:45.000Z (about 2 years ago)
• Default Branch: master
• Last Pushed: 2024-03-24T04:05:27.000Z (about 2 years ago)
• Last Synced: 2024-05-01T11:33:04.429Z (about 2 years ago)
• Topics: ai, android-app, image-classification, onnxruntime, react-native, weeb
• Language: TypeScript
• Homepage:
• Size: 3.87 MB
• Stars: 0
• Watchers: 1
• Forks: 0
• Open Issues: 0
• Metadata Files:
  • Readme: README.md

Awesome Lists containing this project

README

          
# WaifuTagger

![banner](./banner.png)



    





    





    



Run SmilingWolfs [WD Convnext Tagger v3](https://huggingface.co/SmilingWolf/wd-convnext-tagger-v3) model on mobile and [WD SwinV2 Tagger v3](https://huggingface.co/p1atdev/wd-swinv2-tagger-v3-hf) on desktop!

## 🎮 Platforms

Android and Desktop (Windows, MacOS, Linux)

## 💾 Download

Take a look at the latest [release](https://github.com/KuzuLabz/WaifuTagger/releases/latest)!



    



## 🤗 Model

#### Mobile

To run the model on mobile, I quantized it and added preprocessing using [onnxruntime-extentions](https://onnxruntime.ai/docs/extensions/). This model is available on my [HuggingFace](https://huggingface.co/Smashinfries/wd-convnext-tagger-v3-mobile). You can also find more details about the process there.

#### Desktop

Desktop is using p1atdev's transformers version of the [WD SwinV2 Tagger v3](https://huggingface.co/p1atdev/wd-swinv2-tagger-v3-hf) model due to some ops (decodeImage) not working with onnxruntime-web. 

The WASM backend is used for inference.

## 🎯 Model Accuracy

#### Mobile

If you get bad results, it's highly likely that I did something wrong with the preprocessing. Any tips for improvement would be appreciated.

#### iOS Notes

- This app uses [expo-share-intent](https://github.com/achorein/expo-share-intent) and will need additional setup for iOS.