Live demonstration: Real-time implementation of a proto-object-based dynamic visual saliency model

Visual saliency is an important aspect of the human visual system, as it allows us to process an overwhelming amount of visual information in real-time. Rather than processing the entire visual field in parallel, we focus our attention on only interesting regions for performing higher levels of processing. There are many approaches to modeling visual saliency given visual information, however, few models have been implemented for real-time processing and real-world applications. This demonstration takes the proto-object-based dynamic visual saliency model one step further by implementing it on an FPGA (Field Programmable Gate Array) for realtime processing. This allows for many interesting applications including those relating to augmented vision and autonomous navigation using visual processing.