A Look At Lobe – A Free, Easy-To-Use Tool For Training Machine Learning Models

A few months ago I heard about a new tool from Microsoft called Lobe which makes it easy to train machine learning models. It’s nothing to do with Power BI but I find anything to do with self-service data analytics interesting, and when I finally got round to playing with it today I thought it was so much fun that it deserved a blog post.

You can download it and learn more at https://www.lobe.ai/ and there’s a great ten minute video describing how to use it here:

The most impressive thing about it is not what it does but how it does it: a lot of tools claim to make machine learning easy for non-technical users but Lobe really is easy to use. My AI/ML knowledge is very basic but I got up and running with it extremely quickly.

To test it out I downloaded lots of pictures of English churches and trained a model to detect whether the church had a tower or a spire. After I labelled the pictures appropriately:

…Lobe was able to train the model:

I could test it inside the tool. The model was able to tell whether a church had a tower:

…or a spire:

…very reliably!

If I have one criticism it’s that when you want to use your model things get a lot more technical, at least compared to something like AI Builder for Power Apps and Power Automate, but I guess that’s because it is just a tool for training models. There have been some recent improvements here though (see this blog post) and Lobe does provide a local API for testing purposes that can be consumed in Power BI with some custom M code.

Here’s an example of how to call the local API in Power Query:

let
  Source = Folder.Files("C:\Churches"),
  #"Removed Other Columns"
    = Table.SelectColumns(
    Source,
    {"Content", "Name"}
  ),
  #"Added Custom" = Table.AddColumn(
    #"Removed Other Columns",
    "CallAPI",
    each Text.FromBinary(
      Web.Contents(

        //Insert Lobe Connect URL here                              
        "http://localhost...",
        [
          Content = Json.FromValue(
            [
              image = Binary.ToText(
                [Content],
                BinaryEncoding.Base64
              )
            ]
          ),
          Headers = [
            #"Content-Type"
              = "application/json"
          ]
        ]
      )
    )
  ),
  #"Parsed JSON"
    = Table.TransformColumns(
    #"Added Custom",
    {{"CallAPI", Json.Document}}
  ),
  #"Expanded CallAPI"
    = Table.ExpandRecordColumn(
    #"Parsed JSON",
    "CallAPI",
    {"predictions"},
    {"predictions"}
  ),
  #"Expanded predictions"
    = Table.ExpandListColumn(
    #"Expanded CallAPI",
    "predictions"
  ),
  #"Expanded predictions1"
    = Table.ExpandRecordColumn(
    #"Expanded predictions",
    "predictions",
    {"label", "confidence"},
    {"label", "confidence"}
  ),
  #"Pivoted Column" = Table.Pivot(
    #"Expanded predictions1",
    List.Distinct(
      #"Expanded predictions1"[label]
    ),
    "label",
    "confidence",
    List.Sum
  ),
  #"Changed Type"
    = Table.TransformColumnTypes(
    #"Pivoted Column",
    {
      {"Tower", type number},
      {"Spire", type number}
    }
  ),
  #"Removed Columns"
    = Table.RemoveColumns(
    #"Changed Type",
    {"Content"}
  )
in
  #"Removed Columns"

You can export models to a variety of other places for production use, including Azure Functions and Azure Machine Learning.

Definitely something to keep an eye on, especially because it will soon be able to do object detection and data classification as well as image classification.