Licence
AI Hub runs open-source machine-learning models and frameworks entirely in your browser. This page lists attribution and terms for the libraries, models, and datasets used.
Frameworks
AI Hub runs entirely in your browser on Transformers.js (Hugging Face), TensorFlow.js, Tesseract.js, and ZXing — open-source under the Apache 2.0 licence — together with node-qrcode and JSZip, which are MIT-licensed. — Transformers.js · TensorFlow.js · Tesseract.js · ZXing · JSZip: https://huggingface.co/docs/transformers.js
Models
Tools use open models hosted on Hugging Face — including ViT-GPT2 (captioning), U-2-Net (background removal), MobileNet V2 (image classification), Depth Anything V2 (depth), YOLOS (object detection), Swin2SR (upscaling), CLIP (zero-shot classification), SlimSAM (segmentation), DistilBERT (sentiment & question answering), BERT-NER (named-entity recognition), Kokoro (text-to-speech), and DistilBART-CNN (summarisation). Each model is governed by its own licence on its Hugging Face model page. — Hugging Face Models: https://huggingface.co/models
Quick, Draw! Dataset
The Sketch tool's recognition model is built on Google's Quick, Draw! dataset, released under the CC BY 4.0 licence. — Quick, Draw! Dataset (CC BY 4.0): https://quickdraw.withgoogle.com/data
Changes
This page will be updated as new models or datasets are added.