Understanding Defect Detection App

A Defect Detection App solution uses an edge device to perform offline analysis and processing of images from a production line, with an Amazon Lookout for Vision model. You use the Defect Detection App Console to train and deploy a model to a station hosted on the edge device. You then use an application (Defect Detection Station App) at the station to manage and run workflows. A workflow defines the steps taken to analyze an image and process the analysis results. With a workflow, you can analyze images from cameras on the station, or with images retrieved from a folder on the edge device. You can automate image analysis by using a digital signal input trigger to run a workflow, or run the workflow manually from the Station App. The analysis results are saved to a local folder on the edge device. Optionally, you can send digital signals based on the results of the prediction that the model makes.

Important

Before you can use Defect Detection App, you need your tenant sign in details. For more information, see Signing into the Defect Detection App Console.

Set up the edge device

An edge device that you use with Defect Detection App is an Industrial PC (IPC) that hosts a Defect Detection App station. The edge device is located where you need to analyze images, such as a production line where you want to find manufacturing flaws on completed circuit boards. The edge device has one or more attached cameras that provide images for analysis with the model.

You create a station by using the Defect Detection App Console. When you create a station, you specify the number of workflows that you want your station to have.

Important

You are charged a monthly subscription for each active workflow.

As part of station creation, you deploy the station software to your edge device. Once completed, you can use the Defect Detection App Console to deploy models to the station. You also need to set up the image source on the station. An image source is the location where a workflow gets images from. An image source can be a camera on the edge device, or a folder on the edge device. You set up the image source using the Defect Detection Station App. If the image source is a camera, you also use the Station App to configure the camera settings, such as the exposure time for image capture.

More information: Setting up a station

Train and deploy an Amazon Lookout for Vision model

An Amazon Lookout for Vision model is a machine learning model that can find anomalies in images. You can train an Amazon Lookout for Vision model to classify images as normal or anomalous. You can also train a model to find the location of anomalies on an image (segmentation).

You use the Defect Detection App Console to train and manage the Amazon Lookout for Vision models that you deploy an edge device.

You train the model with a dataset of images that you supply and label. You can capture images for your dataset by using the Station App.

After you train a model, you can deploy the model to a station on the desired edge device.

More information: Training and deploying a model

Running a workflow

You use a workflow to analyze an image and process the analysis results. The workflow steps are:

• Getting an image from an image source. An image source can be a camera on the same network as the edge device or an image folder on the edge device.

• Analyzing the image for anomalies with a model that you create. You can manually analyze an image or use a digital signal to trigger the running of a workflow. If the image source is a camera, the workflow captures an image and then analyzes the image. If the image source is a folder, the workflow analyzes the oldest image in the folder that it hasn't yet analyzed.

• Performing output tasks based on the analysis results. For example, you can trigger an output device if the model predicts an anomaly within an image.

You specify the number of workflows you need for a station when you create or edit a station. You use the Station App on the station to configure the workflow settings, such as setting the image source or setting a digital signal that triggers the running of a workflow.

When a workflow runs, Station App stores the results on the edge device for later use. You can use the Station App or API to view the results for the last analyzed image.

More information: Analyzing images with the Defect Detection Station App