Computer Vision News - February 2018

as a catchphrase. Always take the most pragmatic position. Insisting on deep learning models in every phase of our pipeline is not something that has paid rich dividends. That is something that I would like to share with your readers. ” In case of false positives, when the system has not identified the right person properly, they have incorporated a cross-function into their pipeline: they developed the system taking into account the user experience, the UI/UX , especially for a person in desperate need of authentication. Today, UnifyID has a budget for UI/UX and they have conducted studies about it. They collected huge amounts of data from people of all different backgrounds. They studied how to set the hyper-parameters of their machine learning system in order to get an optimum point. Then they could send the hyper-parameters in such a way that minimizes the false positive as much as possible . They also provide a hybrid mechanism, using fallback authentication techniques . They provide a seamless fallback on active authentication which can be as simple as asking a question. In some cases, their partners insisted on having the option to revert to passwords . Through the development of their system, they learned that many people do not want to abandon the use of passwords. It’s part of their routine. When they feel it is seamlessly authenticated, they become very nervous. They wonder how the technology knows enough to identify them. As a consequence, UnifyID also allows for elegant fall back to active authentication techniques . Computer Vision News 21 UnifyID Application Accelerometric data activity dimensionality reduction