Quality above all!

One of the easiest ways to demonstrate the power of an application is to show the customer how it works live, with real-time data, what better than Visual AI.

IBM Maximo Visual Inspection is composed of key components:

• Datasets
• Models:
  • Image Classification
  • Object Detection
  • Video Action
• Projects
• API Configuration for IBM MVI Mobile Edge.

I'll show you how to prepare a beautiful demo.

Some definitions first:

IBM Maximo Visual Inspection is composed of fundamental parts:

I'll show you how to prepare a beautiful demo.
Some definitions first.

• Dataset: There are two types, training and inspection. Training allows you to label and classify images or video frames that contain what you're looking for. Inspection is for when you create a project and put it to work through the Edge barrier, whether mobile or server-side.
• Models: There are various types of models to train, it all depends on what you're looking for. I recommend studying the visual characteristics of what to look for in the image capture from future cameras to be viewed in the inspection.
  • Image classification: Classifies the entire image as a general label, for example: an empty conference room or a full conference room.
  • Object detection: Detects objects within a frame, like the quality example I'll share with you.
  • Video action: Labels frame captures from a video, for example, from minute 0:01 to minute 0:02, something happens in the video, such as a fall, someone taking off their helmet, etc.
• Projects: here you simply indicate which model you will use with which inspection dataset, which you will later retrieve from the mobile app or MVI Edge server.

1. The first file is the already labeled dataset, automotive leather stitching.

The dataset contains 1,102 images labeled with non-conformities of automotive leather stitching, with a single label "non-conformities." You're welcome, this is the hardest job.

2. The second file corresponds to the trained model.

This corresponds to a Tiny Yolo model, which will allow you to download it to an MVI Edge Mobile. To do this, download the app to an IBM Visual Inspection iOS device. Follow the steps to configure your Edge environment (scan the API QR code in IBM MVI Core). Make sure you have created the project for both the inspection dataset and the deployed model.

Once downloaded to the mobile app, you can configure the inference to fail on video.

After that, just enable streaming mode and voila!.

3. I'm leaving you some images so you can try the model in the mobile app.

To surprise your client, use the MVI mobile app and then put the images I shared with you on a screen and start analyzing them in real time. You can move the images and see how quickly MVI Edge mobile analyzes them.

This should be seen by your client in real time as you scroll through the images.

If you ever use it, at least mention me hahaha, 

Tell me if it worked for you, I wish you the best.

David Iñiguez.