nsfw_data_scraper
Collection of scripts to aggregate image data for the purposes of training an NSFW Image Classifier
- 13k
- Shell
- MIT
- 3 y ago
Collection of scripts to aggregate image data for the purposes of training an NSFW Image Classifier
Deep learning based content moderation from text, audio, video & image input modalities.
Alignment-research scaffold (autoresearch-style) for LLM guardrails: search over a single policy.md surface
Guardrail capabilities for Pydantic AI — cost tracking, prompt injection detection, PII filtering, secret redaction, tool permissions, and async guardrails. Built on pydantic-ai's native capabilities API.
Self-hosted content moderation API that outperforms Amazon Comprehend. 100% offline, your data never leaves your server. Text + Image moderation.
Local decision model with calibrated probabilities: send a state and yes/no, choice or score questions, get a probability for every option. 0.8B GGUF on CPU, Jev-style API.
AI Firewall & LLM security toolkit - protect your AI applications from prompt injection, jailbreaks, PII leakage, and adversarial attacks
Free, self-hosted image moderation API in Go. Check NSFW images on your own server with HTTP, gRPC, Docker, and optional OCR. No subscription or per-image API fees.
A framing proposal and movement site on engagement feeds and minors: regulate the loop, re-attach consent, de-amplify don't censor. The named standard is brake integrity.
Open-source, self-hostable community social network: feeds, communities, real-time messaging, events and AI-assisted moderation. Terraform for AWS, Floci for local.
Moral Spectrum Analyzer — measures the moral coverage of AI content-evaluation, moderates where validated, escalates where not, and discovers missing dimensions. GTC 2026 prototype (Pillar A).
Dataset of public interest interventions on Twitter for politicians and candidates during the 2020 US General Elections
Geometric deep learning for content moderation — hyperbolic policy embeddings, manifold decision boundaries, and distributional fairness
Node.js + Express Backend for UltimateHealth – Secure REST APIs for multilingual health articles, podcasts, user reviews, moderation, JWT auth, and admin analytics. Built with MongoDB + Swagger documentation.
A harassment shield for LinkedIn DMs: abuse hides behind a warning, and a sender's messages become a PDF evidence report. A Chrome extension (Plasmo, MV3) on an Express API and a Gemini moderation service with failover. 1st at HackWIE 3.0 and Code Kshetra 2.0.
Reference platform for image, audio, and video analysis using FastAPI, vision models, Whisper, moderation workflows, and a web interface.
Text classification model, used with transformers.
Increase trust in your wiki knowledge base via approval workflows
Text classification model, used with transformers.
Admin dashboard for UltimateHealth — ensuring accurate, original, and high-quality health content through smart moderation, collaborative workflows, and automated checks.
OneDecision-VisionGuard-Demo is an interactive, visual content-moderation and guardrail workflow powered by the prithivMLmods/OneDecision-VisionGuard-9B-SFT vision-language classification model.
One-click NSFW detection + censoring (mosaic/blur/bars/manga/glitch, brush mask, stylize, JP export presets) for Stable Diffusion WebUI Forge. NudeNet-powered.
Real-time AI / solicitation / parrot-bot scores for X posts and replies. Chrome extension + Workers API on open-compute, judged by TypeSafe Jev.
Library of 29 custom topic guardrails for Prisma AIRS Runtime Security with deploy scripts and industry profile templates
MCP server that gives AI agents typed decisions instead of prose: classify, route, score, and flag with calibrated probabilities. npx -y openjev-mcp
Self-hosted NSFW video moderation pipeline: NudeNet (local ONNX) classifies ffmpeg-sampled frames, flags unsafe content, and routes it to a human-in-the-loop review queue — every video, frame crop, and report stored on Backblaze B2 via the S3-compatible API. Next.js + FastAPI.
Text classification model, used with transformers.
Official ZeroGPU nodes for n8n. ZeroGPU is a compute-efficient inference provider for apps and agents — purpose-built small and nano language models on an edge network that run the repeatable tasks frontier models shouldn’t, ~10x faster and 50%+ cheaper. Auto-scaling, with zero GPU infrastructure. Plug in and you’re live.
Interactive Shieldstral policy graph explorer
Runs in the browser on Hugging Face
Experimental SFW–NSFW classifier for video, image and text using OpenRouter's Decisions API and a rule-based S0–S5 scale. Research prototype, not a moderation system.
A high-performance multi-provider LLM API gateway in pure Go. Unified OpenAI-compatible entrance with protocol translation (Anthropic/Gemini), load balancing, circuit breaking, quotas, bypass-resistant content moderation, and an embedded web console - all in a single static binary.
Successfully built and deployed a production-ready AI media processing microservice featuring asynchronous BullMQ queues, MinIO storage, and multi-model AI pipelines for image captioning, object detection, and content safety verification.
End-to-end responsible-AI audit of a DistilBERT toxicity classifier: identity-group bias audit, adversarial attacks, mitigation techniques, and a production guardrail pipeline. Built on the Jigsaw Unintended Bias in Toxicity Classification dataset.