PR2 Robot Learns To Read Text With Different Text Sizes & Colors

The University of Pennsylvania's GRASP Lab has made their Robot PR2, who is nicknamed 'Graspy', to read text. So now, it can roam about the premises and read aloud the various notice boards, signs and warnings.

Have you ever wondered how we started reading text as a toddler? It was and is a two step process: 1. Watch 2. Decode. We recognize words their shape and format and then by analogizing and attaching the pronounceable word parts and then read them. We analogize the word by recalling some word with the same spelling pattern and make the new unfamiliar word rhyme with the word we remember.

#-Link-Snipped-#

Teaching a robot to do that, is simply a dexterous task. Especially with different fonts and styles that we use to write the words, the process of teaching a robot becomes more complex.

The approach that PR2 uses for recognizing and reading the words is as follows:

1. Capture Image Frame

2. Detect Text Region

3. Crop Out Text Region

4. Perform Text Recognition

What will interest you more is that this project will be #-Link-Snipped-# for both PR2s and generalized ROS platforms so that you can enhance the capabilities of your robot within minutes.

Check out the demonstration of PR2 Robot aka Graspy reading:



Source: #-Link-Snipped-#