NSFWJSWrapper

A NSFW & SFW detecting demo for iOS , based on nsfwjs and assembled via WKWebKit

Feature

iOS side indecent content checking，like Porn/Neutral/Sexy/Hentai/Drawing

1. git checkout main for iOS hosted an GCDWebServer serving the tensorflow models and required js loading, off-line supported, usually takes 13-27 seconds to initialization.
2. git checkout load_from_remote,within the file of NSFWJSWrapper/nsfwjs/src/NSFWJSMessager.html, replacing the configs tensorflow_url,nsfwjs_url, model_url(optional) , to use the function user should be on-line, required js files tensorflow models should be deployed to a remote server，usually for team work use case.

Next Step

• single/multipule UIImage(s) detecting base on tensorflow

• [-] load tensflow.js / nsfwjs.js from bundle to speed up the initialzation native self hosted http server

• [-] UGC contents represented in other way: character sentences, voices detecting feature

• iOS/Android unified interface integration, under most developing cases should be like this

PS: There may be Not Safe For Work contents in the Demo

Reference:

一个短视频APP视频内容安全审核的思路调研及实现汇总

色情图片离线识别，基于TensorFlow实现。识别只需20ms,可断网测试，成功率99%

较成熟方案：

open_nsfw_android: open_nsfw_android

open_nsfw_ios: open_nsfw_ios