Create a Project for Real-Time Image Classification on Mobile Devices
Question
You plan to use Azure Custom Vision service to classify images.
The state of art machine learning algorithms offered in Azure through this service saves you time and effort.
To create a classifier in the Custom Vision web page, you are required to create a project.
Select the Domain that you will choose for creating the project that meets following requirements: Allows models to be exported and run locally Optimized for real-time classification on mobile devices.
Answers
Explanations
Click on the arrows to vote for the correct answer
A. B. C. D.Correct Answer: B.
Option A is incorrect because the General [A1] domain is used for large datasets.
It offers better accuracy than the General domain.
However, this domain is not purposed for exporting the model.
In order to do so, you would need to first convert the domain to a compact domain.
Option B is correct because the General (compact) is intended and optimized for real-time classification on mobile devices.
The models might be slightly inaccurate but you can also export them and run them locally for real time classifications.
Option C is incorrect because the General [A2] domain offers better accuracy than General and General [A1] domains.
However, this domain is not purposed for exporting the model.
In order to do so, you would need to first convert the domain to a compact domain.
Option D is incorrect because the General domain is optimized for the general image classification tasks.
It serves as the best choice when you run your models in Azure and do not know which domain to choose from.
Reference:
To learn more about selecting a domain in custom vision service, use the link given below:
To meet the requirements of exporting and running the model locally and optimizing it for real-time classification on mobile devices, the best option would be to choose a compact domain.
The compact domain is optimized for small models, making it ideal for running on mobile devices. It is specifically designed to be lightweight and efficient, allowing for fast and accurate real-time classification on mobile devices.
In addition, the compact domain also allows models to be exported and run locally. This is because it uses a standard ONNX format that can be easily deployed to a variety of platforms, including mobile devices.
Therefore, the best option to choose for creating a project that meets the given requirements would be a General (compact) domain.
