← 开源
slavabarkov

tidy

Offline semantic Text-to-Image and Image-to-Image search on Android powered by quantized state-of-the-art vision-language pretrained CLIP model and ONNX Runtime inference engine

ApplicationsData organizationKotlin
在 GitHub 打开
增长势头
+124 小时新增 Star+0.2%
610
Star
48
Fork
+2
本周
1
贡献者
创建于 2023-02-24 · 更新于 2026-10-15 · 今日第 6609 名
主要开发者
slavabarkov

slavabarkov/tidy 排名第 6,609,有 610 个星标。

同类别比较

slavabarkov/tidy 排名第 6,609,此处有 8 个同类条目;juntaosun/ComeCut 排名第 6,614。

同类别比较
名称排名指标
sxwangsxwang1/chatpassport6,576
语言:
TypeScript
星标数:
625
分叉数:
5
anhaidgroup/deepmatcher6,577
语言:
Python
星标数:
624
分叉数:
130
简介:
Python package for performing Entity and Text Matching using Deep Learning.
Haleclipse/CometixCode6,580
语言:
Rust
星标数:
623
分叉数:
37
简介:
An unofficial Rust reimplementation of Anthropic's Claude Code terminal UI, built on iocraft
checkmarble/marble6,596
语言:
HTML
星标数:
616
分叉数:
90
简介:
Marble - the real time decision engine for fraud and AML
juntaosun/ComeCut6,614
语言:
HTML
星标数:
608
分叉数:
77
简介:
「来剪」轻量级视频编辑器。网页版、桌面版等均可免费使用,功能灵感源自 CapCut 等编辑器。A Lightweight Video Editor. Free for the web, desktop, and more, with features inspired by editors like CapCut.
BurhanUlTayyab/GPTZero6,615
语言:
Python
星标数:
607
分叉数:
92
简介:
An open-source implementation of GPTZero
joeynyc/hermes-skins6,623
语言:
Python
星标数:
605
分叉数:
35
简介:
Custom skins (visual themes) for the Hermes CLI agent
barefootford/buttercut6,624
语言:
Ruby
星标数:
604
分叉数:
98
简介:
Edit Video with Claude Code
关键参数与规格
关键参数与规格
参数值
语言Kotlin
类别Applications
子类别Data organization
关键词Image search, LocalAI
星标数610
分叉数48
贡献者1
创建日期2023-02-24
更新日期2026-10-15
1 天星标数1
7 天星标数2
排名6,609
上次排名15,951
排名变化9,342
7 天星标排名8,732
上次 7 天星标排名11,233
7 天星标排名变化2,501
总星标排名16,557
上次总星标排名16,581
总星标排名变化24
README 摘录

TIDY - Text-to-Image Discovery Offline semantic Text-to-Image and Image-to-Image search on your Android phone! Powered by quantized state-of-the-art large-scale vision-language pretrained CLIP model and ONNX Runtime inference engine. Approach TIDY uses OpenAI CLIP (Contrastive Language-Image Pre-Training) model - a neural network trained on a variety of vision-language pairs. CLIP efficiently learns visual concepts from natural language supervision, which allows TIDY to use it for Text-to-Image retrieval

. CLIP can also be used to get high-quality image representations with high universality and generalization ability providing great results in the Image-to-Image retrieval task. The model used in TIDY is based on open source CLIP implementation OpenCLIP pretrained on LAION-2B, a ~2B sample subset of LAION-5B dataset with english captions. This indexing process may take some time, but it's a one-time event. Once this initial indexing process is complete, the app will store the

index on your device, and any new photos you add to your photo library will be automatically added to the index on the subsequent app launches. Privacy and Security TIDY works entirely offline, ensuring your privacy and security are never compromised. None of your data or images are ever uploaded to a remote server or shared with third parties, ensuring your personal information stays safe and secure. It also means that you can use it anytime, anywhere, even in areas

with poor or no internet connectivity. Text-to-Image Search Video demonstration Simply type in a description of the image you are looking for, and TIDY will retrieve the most relevant matches from your local image library. Text-to-Image search functionality in TIDY goes beyond traditional keyword-based searches! You can use longer and more detailed descriptions to find the exact image you have in mind. Image-to-Image Search Video demonstration Search for visually similar images by choosing

a photo from your device's gallery. TIDY will analyze the image and retrieve images with similar visual features, allowing you to explore and discover new images in a whole new way.

常见问题

记录了多少星标和分叉?

slavabarkov/tidy 有 610 个星标和 48 个分叉。

仓库排名是多少?

slavabarkov/tidy 排名第 6,609。

记录的语言和类别是什么?

slavabarkov/tidy 有 610 个星标,语言为 Kotlin,类别为 Applications。

哪个仓库排名最接近?

slavabarkov/tidy 与 juntaosun/ComeCut 比较,后者排名第 6,614,有 608 个星标。

README

TIDY - Text-to-Image Discovery

Offline semantic Text-to-Image and Image-to-Image search on your Android phone! Powered by quantized state-of-the-art large-scale vision-language pretrained CLIP model and ONNX Runtime inference engine.

Approach

TIDY uses OpenAI CLIP (Contrastive Language-Image Pre-Training) model - a neural network trained on a variety of vision-language pairs. CLIP efficiently learns visual concepts from natural language supervision, which allows TIDY to use it for Text-to-Image retrieval. CLIP can also be used to get high-quality image representations with high universality and generalization ability providing great results in the Image-to-Image retrieval task.

The model used in TIDY is based on open source CLIP implementation OpenCLIP pretrained on LAION-2B, a ~2B sample subset of LAION-5B dataset with english captions.

Image Credit: https://github.com/openai/CLIP

Features and Usage

First Launch

During the first launch TIDY will need to scan through your photo library and create an index of your images. This indexing process may take some time, but it's a one-time event. Once this initial indexing process is complete, the app will store the index on your device, and any new photos you add to your photo library will be automatically added to the index on the subsequent app launches.

Privacy and Security

TIDY works entirely offline, ensuring your privacy and security are never compromised. None of your data or images are ever uploaded to a remote server or shared with third parties, ensuring your personal information stays safe and secure. It also means that you can use it anytime, anywhere, even in areas with poor or no internet connectivity.

Text-to-Image Search

Video demonstration

Simply type in a description of the image you are looking for, and TIDY will retrieve the most relevant matches from your local image library. Text-to-Image search functionality in TIDY goes beyond traditional keyword-based searches! You can use longer and more detailed descriptions to find the exact image you have in mind.

Image-to-Image Search

Video demonstration

Search for visually similar images by choosing a photo from your device's gallery. TIDY will analyze the image and retrieve images with similar visual features, allowing you to explore and discover new images in a whole new way.

Citation

@Misc{tidy,
  title =        {TIDY (Text-to-Image Discovery): Offline Semantic Text-to-Image and Image-to-Image Search on Android Powered by the Vision-Language Pretrained CLIP Model.},
  author =       {Viacheslav Barkov},
  howpublished = {\url{https://github.com/slavabarkov/tidy}},
  year =         {2023}
}
排名相邻
同类项目