New image recognition example added to 2.3
I just added an image recognition example. It is actually script driven and is fairly customizeable, even as a simple example. It shows how to use the Image dataset. Below is an example script that I used. I put in every USA coin type and the two "whatis" commands at the bottom attempt to recognize both a coin it was trained on, as well as a new one.
CreateTraining: width:16,height:16,type:RGB
Input: image:./coins/dime.png, identity:dime
Input: image:./coins/dollar.png, identity:dollar
Input: image:./coins/half.png, identity:half dollar
Input: image:./coins/nickle.png, identity:nickle
Input: image:./coins/penny.png, identity:penny
Input: image:./coins/quarter.png, identity:quarter
Network: hidden1:100, hidden2:0
Train: Mode:console, Minutes:1, StrategyError:0.25, StrategyCycles:50
Whatis: image:./coins/dime.png
Whatis: image:./coins/half.png
Whatis: image:./coins/testcoin.png
Jeff, I know one chapter in the book is on images. Would this example be of any help?




That is a pretty nice example! I will make use of it for my image chapter. I was just starting to think of what example to write, but this will work fine. I trained it with your coin images, and it recognized the nickle just fine. Very nice! The script is a nice touch too, as it allows the example to be used with any sort of image.