In this project you have all the required files and links in order to train Watson Visual Recognition to recognize Halusky In the folders above you can see examples of halusky and sets of examples of items which can be mistakenly categorized as halusky, hence with the negative examples you can point Watson Visual recognition to the correct set of features for real halusky.
- Get and account in IBM Cloud
- Setup istance of Watson Visual Recognition service
- Create a class for Halusky
- As you can see in the dashboard there's already a "NotClass" provided for you
- Attach uploaded files to respective classes (Halusky to halusky class, anything else to NotClass)
- Start the training by clicking "Train" button
- After some time you will have a class trained to recognize halusky.
Now you can utilize REST endpoint or CoreML to use your trained class.
Requirements:
In the files of this repository you can find also a flow designed to utilize your trained class and a Telegram bot which you can create via talking to BotFather in Telegram platform.
As you can see in the visual recognition service all you need to start a class is 10 pictures, please do keep in mind that the number of example (positive & negative) images is important as training events count towards your API usage which with lite plan is limited.