Mini toolsNSFW Model
By Infinite Red
An image classification model for identifying not-safe-for-work content. It sorts images into five categories (drawings, hentai, neutral, porn, sexy) and returns probability scores for each. Available in Keras, TensorFlow, and TensorFlowJS formats, with both Inception V3 (299x299) and MobileNet V2 (224x224, including a quantized variant) versions. Achieves 93% accuracy per its published confusion matrix. Usable via Python library, command line, or JavaScript (powers the NSFWJS project). Intended for content moderation use cases.
Image
Gen 4
Overview
Image classification model trained on 60+ gigs of data to detect NSFW content. Classifies images into five categories: drawings, hentai, neutral, porn, and sexy. Built on Inception V3 and MobileNet V2 architectures with 93% accuracy.
Last updated: July 6, 2026
