Demo: Custom Vision Service

The video tutorial covers the use of custom vision service, focusing on creating and using cognitive services for image classification. It demonstrates how to train a model with images of apples, bananas, and oranges, and tests the model's accuracy. The tutorial also explains the process of publishing the model and using it in applications. Additionally, it introduces object detection as another feature of custom vision.

Why might you need to create a custom vision model instead of using a predefined one?

• To save time and resources

• To address specific application needs

• To use more general data

• To avoid using cloud services

What is the first step in setting up a Cognitive Service for image classification?

• Creating a new project in Custom Vision AI

• Selecting a subscription and resource group

• Uploading images to the service

• Testing the model with sample images

What is the purpose of tagging images in Custom Vision AI?

• To reduce the size of the dataset

• To categorize images for model training

• To encrypt the images

• To enhance image quality

What should you do if your specific domain is not available in Custom Vision?

• Use a different service

• Skip the domain selection

• Create a new domain

• Choose a general domain

What indicates a well-trained model in the performance matrices?

• No accuracy percentage

• Low accuracy percentage

• High accuracy percentage

• Random accuracy percentage

What information is needed to use a published model?

• Training data and results

• Image tags and categories

• Project ID, key, and endpoint

• Model name and description

What is the role of the endpoint in using a cognitive service?

• It is used to store data

• It represents the service's address

• It acts as a security measure

• It provides the service's physical location

How can you test the accuracy of a trained model?

• By using a quick test with sample images

• By checking the model's code

• By reviewing the training data

• By comparing it with other models

What is the main difference between classification and object detection in Custom Vision?

• Classification identifies categories, object detection locates objects

• Classification is faster than object detection

• Object detection uses more data than classification

• Object detection is more accurate than classification

What is the first step in creating a new project for object detection?

• Uploading images

• Tagging images

• Selecting a project type

• Creating a new project