Optical Character Recognition

Optical Character Recognition (OCR) is used extensively for collecting data into your Private Cyberspace. Machine Learning (ML) has made OCR more accurate than ever before, your Private Cyberspace has made ML more secure than ever before.

Unlike other Artificial Intelligence (AI) systems, we do NOT use your data to train our AI, we let you train your own AI. the AI nodes actually runs on your own devices or on devices you have control of.

1. OCR Node Types

You can access advanced OCR directkly on the 88.io web console.


1.1. Basic OCR Node

All Private Cyberspaces come with a Basic OCR Node that can handle most daily OCR use cases:

  1. recognising your weight displayed on the scale
  2. recognising the registration number plate of the taxi that picked up you children
  3. recognising the name of the restaurant you are about to enter

Basic OCR Nodes do NOT need special AI acceleration hardware (e.g.GPU, TPU).

1.2. Advanced OCR Node

Some more complex OCR situations (e.g. scanning of receipts) currently requires hardware acceleration (CUDA 10.1 or above). This is delivered through Advanced OCR Nodes.

Some receipts are challenging even for humans to read, so it can be tough for computers. But having the productivity gain of having AI to read receipts is too much to ignore, so receipt reading a major focus of Private Cyberspace development.

A good introduction is here:

2. Training

If you found a Node that works well for a specific situation - like scanning the receipt from a certain supermarket - share (or may be even sell) that knowledge!

You do not need to be a PhD not even need to understand how OCR works, you just need to tell others which Node recognises text from images the best in your specific situation - anyone can participate!