Custom Vision

Initial Setup: Create a New Project & Resource Multi-Label Classification

Once a Microsoft Custom Vision model is trained, it can be used to accurately predict the presence of each fruit type in our various testing images. Let's start by creating a new Custom Vision Resource Group and Project for this example.

Step 1: Navigate to Custom Vision

  • Go to www.customvision.ai and select New Project
  • By Resources click the blue create new hyperlink

Step 2: Resource Name

  • Name: Choose a unique name for your resource (alphanumeric, no spaces or special characters)
Example: Fruit_Classifier

Step 3: Subscription

  • Subscription: Select your active Azure subscription from the dropdown

Step 4: Resource Group

  • Resource Group: Select an existing resource group (such as customvision-rg that we made earlier) or click Create new and name it appropriately

Step 5: Service Kind

  • Kind: Leave this as CognitiveServices - this is correct for Custom Vision

Step 6: Location

  • Location: Choose the region closest to you
Common UK choice: UK South

Step 7: Pricing Tier

  • Pricing Tier: Choose S0 (the standard paid tier)

Step 8: Create Project

  • Once we have created a new resource, fill out the remainder of the form and click Create Project
Custom Vision project setup interface showing the Create new project dialog
Note
We are selecting a Compact model here, as this is both speedy for demonstrations, and makes it possible to export the models we create and train using Custom Vision, for running locally, should we wish to do so.