Polimago: Creating search and classification training sets from one data set


The intention of the sample code of the Search and Classify project is to generate the Polimago search and a classification training sets from just one commonly used data set. This example is a bit complex but I think many users will like the functionality and find this sample useful. The sample has been created using Visual Studio 2015 and Common Vision Blox 13.2. It also uses the open source Newtonsoft JSON component which you will find in the solution root directory.
After generating the classifiers from a data set, the sample also provides functions to test the classifiers intensively. Beside testing the search and classification, it provides functions to generate heat maps that plot the confidence of either the search or the classification task in a pseudo-colour image with red showing high confidences and blue showing low confidences. This gives a good visualization of the quality of the classifiers.
You will find the C# source in the attached ZIP file as well as you will find an application note describing how to use it attached to this post.SearchAndClassify.zip (1.0 MB)
2019-04 SearchAndClassify Application Note.zip (1.2 MB)