A unique feature differentiating intuVision VA from other video analytics is our robust object classification with unrestricted categories. In addition to default person and vehicle classification types, users can add custom object categories and quickly train classification models directly from the intuVision VA Admin Application.

Whether the project requirements call for keeping separate counts for cars, trucks, and motorcycles, or excluding animals from triggering perimeter intrusion alarms, intuVision VA trained classification makes it possible.

The user trainable object classification is part of intuVision VA's patented algorithms that learn scenes and recognize objects as the video is being analyzed, without having to collect large number of model training samples. The moving objects detected and tracked with intuVision VA readily provide the model training samples. Since each object track includes several pose snapshots of an object type, those constitute good training data for building classification models. Just a few track sequences are needed to build a model for an object category and training takes only seconds!

Training object classification from a particular camera view has the advantage of learning objects as they appear in that view for better recognition. Trained models generalize well and can be used in similar camera views. The classification models can also be retrained and updated for improved recognition, as more examples of an object type is observed.

Object classification is used in intuVision VA events to limit alarms to object types such as people lingering around a parking lot, or to count only a specific type of object such as trucks entering a bridge.