Working With AI Studio People & Faces

Creating a Model for Facial Recognition

To enable Machine Learning services to recognize specific people and faces, create a model for facial recognition using the steps below. For more information on training models, refer to People and Faces in Wasabi AiR.

Requirements

• Use one to five clear photos of the same individual.
• Use photos of a relatively high resolution (Min: 800 pixels Max: 1200 pixels).
• If possible, select and upload photos from multiple angles of the person’s face. For example, facing front, side angle, down, and so on.
• Use well-lit photos because they perform the best. Try to avoid shadows or images that are too dark or too bright.
• Choose photos that only have one person in them.
• Avoid using photos of people wearing glasses, hats, or any other face covering in order toachieve the most accurate results.
• Choose photos with different facial expressions. For example, a smile, frown, serious, happy, and so on.

1. On the Navigation Panel, click AI STUDIO. 
2. Click anywhere in the training box to upload one to five images from your local machine. Optionally, drag and drop images into the training box.
3. In the NEW PERSON NAME field, enter a name to identify the person in the images.
   If a person has already been created in Wasabi AiR, the type-ahead will suggest the name in the drop-down.

   

4. Click CREATE NEW PERSON.

   

Once images have been successfully uploaded, the newly created person will appear in the list on the left. Notice the in the upper left of each photo. To remove an uploaded image from the training set, click. 

A message will confirm that the photo was successfully removed from the training set, as shown below.

Uploading Images to a Training Set

Once you have created a model for facial recognition, you can upload additional images to improve the model. 

1. On the Navigation panel, click AI STUDIO.

   

2. Select the model from the list view. Optionally, search the name in the type field.  
3. Click +. 
4. Upload additional images from your local machine. Once complete, changes will be automatically saved.