How to Set Up the EL Classify Tool – Christmas Edition

Cognex really have made it very simple for anyone to setup and classify products using this tool.

With inbuilt AI capabilities, the Cognex In-Sight 2800 camera will have your products classified and ready to automate your production line in no time.

For more on how AI is integrated into machine vision systems, check out our No BS Guide to AI in Machine Vision.

When we classify a product in the machine vision world, we’re teaching the software what a particular product looks like – i.e. linking an image of a product to a name (or classification).

Therefore, when that product is detected by the camera in the future, it will know what it is and be able to carry out your instructions for that product.

For example, you may want the product to be rejected, turned around, picked or measured.

By classifying the products in the software, we’re able to detect that product regardless of how it’s presented to the camera.

Lighting and focus can both be optimised at the click of a button, or if you prefer, they can be manually adjusted using the sliders.

Once lighting and focus are set up, we can move into the Inspect tab and define the area we want the software to detect within the image.

With the product under the camera and the first image created, we can now tell the software which product this is. To do so, we simply give the product a name, so the software knows that when it sees a product like this, it should classify it as that name.

We now need to teach the system to recognise that product in different orientations, so it can build a picture of what to expect when it detects that product.

This step simply involves taking several different images of the product at different orientations, to ensure the software is detecting and classifying the product in the correct way.

Initially, the software won’t detect the product when you change the orientation, but after you give it a few images and tell it how to classify each one, it will quickly learn. You’ll notice how it begins to classify your products correctly.

The software reports a ‘predicted class score’ each time, so you’ll be able to monitor its progress and see how well it’s managing to detect the product over time.

In the case of the Christmas chocolates used in the video below, we also wanted to train the software to classify the chocolate if it were to detect the back of it, rather than the front.  

Now it’s time to repeat the process for all the other products – in this case chocolates – so the software knows what each of them looks like. It will then detect and classify them as they come through the production line.

We’ll add a class for the next product, then repeat the ‘teaching’ process until the software has learned what it is.

Repeat this process for all your products.

Once all your products have been trained into the system, you can set the camera to repeating trigger mode and test it. As you place each product under the camera, you should see how the software detects and classifies each one.

We’ll guide you through the available cameras and software systems to ensure you optimise your setup with the correct equipment.

Book a discovery call now.