Mobile AI Platform
  • DOCUMENTATION
    • Overview
    • Getting Started
      • Project Creation
      • Manage Your Team
      • Mobile AI Data App
    • Zero-Shot Classifier
      • Taxonomy Creation
      • Training & Packaging
        • Preview in Mobile AI Data App
        • Custom SDK Integration
    • Few-Shot Classifier
      • Taxonomy Creation
      • Data Collection
      • Annotation (Optional)
      • Training & Packaging
        • Preview in Mobile AI Data App
        • Custom SDK Integration
    • Standard Classifier
      • Step 1 - Domain Architecture
      • Step 2 - Collection
        • Sources
        • Uploads
        • Uploads in Browser
        • Uploads in Mobile AI Data
      • Step 3 - Annotation
      • Step 4 - Training
        • Create a Dataset
        • Train on a Dataset
      • Step 5 - SDK Packaging
        • Preview in Mobile AI Data App
        • Custom SDK Integration
  • HELP
    • Support
    • Tips
    • Key Concepts
    • FAQ
Powered by GitBook
On this page
  1. DOCUMENTATION
  2. Zero-Shot Classifier

Taxonomy Creation

Define the items you want to recognize

When you click the Taxonomy (first) step for a Zero-Shot model, you will be taken to a special version of the taxonomy page. Here, you can create, edit, delete, or even temporarily disable your labels, as well as immediately live-test the results on your model in the browser.

Now you can create your labels. For zero shot it is important to know:

  • Label Name: This is what the user will see when the item is recognized. This can be any word, and has no impact on the recognition.

  • Description: Visually describe what you want to be recognized and called the Label Name associated with this description

For example, I will add 2 labels to my model, "Person" and "iPhone". These are both quite basic objects and don't need much more description, so I will add them as:

Label Name
Description

iPhone

an iphone

Person

a person

Now, if I select my webcam from the dropdown, I should see that while my face is in frame, it will predict "Person" but if I hold my iPhone in front of the camera, it will switch to iPhone.

If I want a somewhat more specific prediction, I might need to add more detail like the following:

Label Name
Description

Happy Person

a person's face smiling

Person

a person with a neutral face

Sad Person

a person's face frowning

iPhone

an iPhone

Now if I change my facial expression, I will get the appropriate predictions.

You can edit your labels to try new descriptions using the Pencil icon, or delete them using the trash can. If I want to disable a particular label from entering the model, but I don't want to commit to deleting it from my project, I can press the Eye icon to hide the label from use.

If at any point you need help from the Passio team, please reach out to us at support@passiolife.com

PreviousZero-Shot ClassifierNextTraining & Packaging

Last updated 1 year ago