Unanswered

AI Builder versus Custom Vision in Azure

(0) Share

Report

Posted on by mmossel Microsoft Employee

Hi,

We want to build an image recognition model that detects holes in asphalt roads. Imagine a road worker taking a photo to help with the administration that a repair is due, so the main purpose is to make the work of this person more effective and efficient. Ideally, there is an indication of the severity of the hole.

Would it be recommended to use AI Builder, or train this algorithm using Custom Vision in Azure and integrate that with PowerApps? I have not been able to find any documentation to understand if it is the exact same technology running in the background.

Categories:

AI Builder

I have the same question (0)

All responses (2)

Answers (1)

Verified answer

Norah on at

Like (0)

Report

Re: AI Builder versus Custom Vision in Azure

hi @mmossel ,

AI Builder now supports Object Detection, not image classification. And AI Builder is using Azure Custom Vision as the backend.

And based on your description, it seems like you're trying to identify holes in a picture and that should be doable via Object detection model. Overview of the object detection model - AI Builder | Microsoft Docs. The benefit of using AI Builder to create model is you can use the model easily in your PowerApp. We have a dedicated control in PowerApps for Object detection and you just need to add that control in your app and select the model you want to connect to.

But in the case if Object Detection doesn't meet your need, you can definitely train an Image Classification model in Azure and bring it into PowerApps via a custom connector. Custom connectors overview | Microsoft Docs

Thansks
Norah

Was this reply helpful? Yes No
cohenjacobd Microsoft Employee on at

Like (0)

Report

Re: AI Builder versus Custom Vision in Azure

Hi @mmossel, are you still working on the asphalt roads project to detect if there are holes? We are adding support for Image Classification soon via two paths. The first is with "Bring your own Model" where you can train an image classification using tools like Azure ML which is available today. The second is training a model with Lobe, an easy to use Desktop app to train image classification models, and uploading your model AI Builder (coming soon).

Would love to chat more about your scenario if you are still working on this!

Best,
Jake

Was this reply helpful? Yes No