Due to their elastic, viscous or fluid nature, information in highly variant objects can be hidden behind or within themselves. It may also be initially invisible and only recognisable in image sequences.
Our framework incorporates specially designed machine learning algorithms that are invariant to these problems. Within this framework, it is possible to automatically process image data, augment it, inject additional information and train it with the appropriate architecture. Currently, we have extensive data sets and models to recognise the following properties on High Variance Objects: