User Defined Pipeline skills

A User Defined Pipeline (Pipeline) skill allows you to create a sequence of actions to perform on your training data. This provides a customized event detection skill that meets the requirements for your specific use case. For example, you can perform a series of transformations on your training images, identify a region of interest in the images, and then perform an anomaly detection within that region of interest.

A Pipeline skill consists of a series of action blocks. To create your Pipeline, you drag blocks from the Block Definitions toolbar (1 in the following figure) onto the Canvas (2 in the following figure) and then connect them. You can build the Pipeline based on your specific use case and available input images.

Each block in a Pipeline has a unique functionality and performs a specific action on the input, such as transforming the image or generating statistics. The software sequentially processes the blocks, so the output of each block is the input for the next block. In addition to its input and output, each block has a set of configuration parameters (such as threshold values, image refinement properties, or flags for sending data to AVEVA Insight) that you specify to suit your needs.

Certain blocks allow for branching, where you can run two or more independent sets of actions on the same set of inputs. Multi-branching allows you to run multiple parallel scenarios from the same input data. You can then combine the results of the multiple branches into one analytic block. As you build and train the Pipeline, you can tune the elements in the individual branches to achieve the desired results.

The blocks show a preview image or data on the Canvas when you correctly configure the block. The following figure is an example of a Pipeline with preview images and data:

During development, the software restricts the types of blocks that you can connect to each other based on the data types for the input and output. For example, you can't connect a Refine Image block, which outputs a set of transformed images, to a Statistical Prediction block, which requires a set of numeric values as its input.