CogniSight technology
|
CogniSight is a set of visual experts which can be trained to recognize similar or different types of objects for the purpose of a global scene understanding. These experts can be used to monitor known regions of interest, find objects or patterns within frames, track targets, generate hypothesis and more. The learning and recognition mechanism of the experts relies entirely on the CogniMem neural neural marketed by Recognetics. The differentiation between experts resides in the type of objects they are taught, the environmental conditions they are expected to cope with without losing accuracy nor their generalization capability. Experts built their knowledge using a feature which must be a good discriminator for the objects to recognize. In many cases, a single expert cannot be sufficient for the task and needs to consult other experts trained on different features of the same family of objects. A CogniSight engine can synthesize the knowledge built by the neurons into image knowledge files (*.ikf). These files can then be cloned and distributed. They can be simple or complex, public or confidential, free or payable. The key to the success of a CogniSight recognition engine is in its training and validation for a given application. The CogniMem technology is an enabler for advanced supervised and unsupervised learning methods and the Image Knowledge Builder has a long and exiting way to go! |
Download...Presentation
White papers Links and references |