nsfwjs

Why we included this project

NSFWJS gives you a browser-based content filter for user-generated images without the overhead of a separate moderation service. It runs on TensorFlow.js, so classification happens on the client: images never leave the visitor's device and there is no per-request server bill. The pre-trained models return confidence scores across explicit-content classes, and the README cites around 90 percent accuracy on the small model and 93 percent on the mid-sized one, which is honest about its limits. The same library runs in Node.js, so you can reuse it for a backend endpoint or a CI check, and self-hosting the model is simple if you don't want to depend on the hosted copy. If you run a forum, marketplace, or social feed and need a quick, privacy-respecting way to filter uploads, this is the fastest path to a working filter.