Multi-Model PatMax

If the objects you are inspecting have multiple discrete appearance types and the appearances vary significantly and you want to achieve reliable and simple pattern recognition and alignment with a single tool, you can use a PatMax Multi-Model (referred to as Multi-Model hereinafter). That is, the appearance of the objects may fall into clear distinct categories (unlike in the case of the Composite Model where typically the appearance is expected to vary continuously). A Multi-Model tells you which object appearance type(s) it has found in the run-time image and with what alignment result(s). An alignment result includes, for example, the location, pose transform, and score (which is a number between 0.0 and 1.0) for each found instance.

The Multi-Model contains the trained PatMax models you add to it, and it performs pattern alignment and classification on your run-time images based on the added trained PatMax models and the run-time parameters you specify. This means that if you run a Multi-Model on a supplied run-time image containing an object instance, it returns the alignment result from the PatMax model that produces the best result.

The main advantage of using a PatMax Multi-Model over using individual PatMax models is that at run-time, the Multi-Model carries out feature extraction once per run-time image (which features are then input to each model) as opposed to running individual PatMax models which would cause identical feature extraction to occur in every model. See more in Advantage of Using PatMax Multi-Model over Using Individual Models – Pattern Granularity with Multi-Model.

You can supply either standard PatMax models or PatMax Composite Models to the Multi-Model, or the mixture of the two.