ai-training-data

张雪峰语录大全 · 高考志愿填报指南：专业选择避坑、院校推荐、就业前景分析，整理成结构化 JSON 数据集，方便考生家长查阅，也方便 AI 应用直接集成调用。

Local-first Python CLI and MCP server for versioned, cited context from web and document sources.

🚀 Interactive JSONL editor for Claude Code conversation files with real-time file system synchronization. Efficient prompt engineering through conversation editing.

真正的死亡不是肉身的终结，而是被彻底遗忘。主动留下自己，让 AI 记住你，实现数字永生。| True death is not the end of the body — it's being completely forgotten. Leave yourself behind, let AI remember you.

Convert any document format into LLM-ready data format (markdown) with advanced intelligent document processing capabilities powered by pre-trained models.

Open-source Arabic NLP datasets by GDG KSU: collecting, cleaning, and publishing high-quality Arabic and dialectal text data, powered by a modular spaCy pipeline that outputs ML-ready Parquet files for Hugging Face and GitHub.

6,000+ people, 70,000+ images: 10-15 photos per ID (selfies + 2 official ID photos). Perfect for face recognition, KYC verification, identity matching, and biometric training. Ages 18-65, balanced demographics

Local-first desktop application for converting documents into AI training datasets. Generate instruction, conversation, and fine-tuning datasets from PDFs, DOCX, Markdown, HTML, and text using Ollama or cloud LLMs.

Display replay attack dataset for face anti-spoofing and liveness detection. 9,000+ videos from 6,500+ participants across PC monitors and mobile devices

Partial paper mask attack dataset for face anti-spoofing, liveness detection, and presentation attack detection (PAD). 3,000 videos, 50 participants, dual-device capture.

AI 模型训练数据标注规则文档撰写 skill：覆盖文本对话（单/多轮·CoT·RAG·Reward·Agent）与多模态（文生图/视频·VQA·ASR·TTS）共 11 种标注类型

Age estimation face dataset: 10,000+ consented selfies of minors & young adults (10-30 years) with verified per-year age labels. Multi-ethnic, phone-captured. Built for under-18 age gating, age verification, and face recognition

The Metaphysical Theory of Mind for AIs is an open-source protocol that bridges multidimensional human cognition and binary AI processing. By establishing strict logical boundaries between present 'manifest reality' and theoretical 'potential realities,' this framework prevents temporal context bleed and reduces AI hallucinations.

Face recognition dataset with 100,000+ files from 1,000+ individuals. Selfies, videos, and archive photos for age-invariant face matching, KYC, and liveness detection

Silicone mask attack dataset for face anti-spoofing and liveness detection. 12,500+ videos, 18 silicone masks, 40+ accessory combinations. iBeta Level 2 compliant

1,000+ people, 10,000+ files: 8 photos per person + 2 videos

The first open-source, structured, AI-traversable legal knowledge graph of the entire US child protective services system. 2,142 decision nodes across 51 jurisdictions. CC BY-SA 4.0.

Shared IR structs for the North Shore labeling stack (Forge/Anvil/Ingot) — typed datasets, samples, assignments, labels, artifacts, and evaluation runs for labeling workflows

Hierarchical catalog of 1500+ business categories in 21 languages with country-specific localization. JSON, YAML, CSV, Markdown.

Saytica Eval Console—a Next.js, Tailwind v4, and DaisyUI platform for LLM verification and data annotation tracking. Features a high-density leaderboard analyzing model precision metrics (accuracy, latency, cost) and a synchronized, dual-role task board for clients and annotators. Built to elegantly handle incomplete data vectors.