I wanted to use neural networks for pattern matching in c++. The scenario is like this:
The main goal is to determine a product by name when captured by a camera.
A rectangular pack of a product (say for example the container of a toothpaste product) is cut into its edge so that the all of its side are shown in one plane. The camera takes a picture of the pack and compare its patterns to the database.
If the patterns are found from the search, then display the name of the product.
Else, store the patterns of the product to the database with its name (say the brand of the toothpaste).
What I mean by pattern is the distinct feature of the product pack among the other products.
I want to know the following using c/c++ (linux, windows, or mac os doesn’t matter):
- Is there a library that makes work somehow easier?
- If a library is not available, what is the best algorithm you can suggest for pattern matching?
I think first, you will need to do some post processing on picture captured by a camera to normalize it (size, angle, …) For that job, you can use OpenCV.
Then if you want to setup a NN, maybe you should give a try to FANN (Fast Artificial Neural Network) http://leenissen.dk/fann/wp/
The library is compatible with Linux/Windows and really easy to use!