Analytically Defined Spatiotemporal ConvNets for Spacetime Image Understanding

This dissertation introduces a novel hierarchical spatiotemporal orientation representation for spacetime image analysis. This representation is designed to combine the benefits of the multilayer architecture of Convolutional Networks (ConvNets) and a more controlled approach to spacetime analysis. A distinguishing aspect of the approach is that unlike most contemporary convolutional networks no learning is involved; rather, all design decisions are specified analytically with theoretical motivations. This approach makes it possible to understand what information is being extracted at each stage and layer of processing as well as to minimize heuristic choices in design. Another key aspect of the network is its recurrent nature, whereby the output of each layer of processing feeds back to the input. The multilayer architecture that results systematically reveals hierarchical image structure in terms of multiscale, multiorientation properties of visual spacetime. To illustrate the utility of the proposed research, the designed networks has been tested on two spacetime image understanding tasks, dynamic texture recognition and video object segmentation. Further, the role of learning in the context of the proposed analytic approach to network design is systematically explored, thereby yielding a promising hybrid architecture. Finally, a new, large scale dynamic texture dataset is introduced and used for evaluation.