Pipes

Overview

Creating and deploying functionality with ezML is done through the creation of an Application.

At the center of computer vision functionality created with ezML are the pipes. An application can contain multiple pipes, each of which defines a pipeline of computer vision functionality.

The Pipes

A pipe is made up of Layers. Each layer provides a piece of computer vision functionality -- we have a large library of specialized, configurable layers that are open for use.

Each layer in a pipe passes it's result down to the next layer. The next layer would be ran on each output of the previous one

To help you understand how pipes work, let's look at an example.

License Plate Detection

Case: We want to detect and recognize license plates from a stream of dashcam footage

Approach: Building an ezML pipe

• Begin by adding a layer to spot individual cars. Click on 'Add Layer', select the 'General Object Detection' model, and set it to detect 'Car'.
• The next layer focuses on each detected car to locate the license plate. Add another 'General Object Detection' layer, this time configured to detect 'License Plate'.
• To extract the text from each identified license plate, add an 'OCR' (Optical Character Recognition) layer.

The output should now be the location of each car, license plate, and respective license plate text in the image,