Computer Vision News - July 2016

Our CEO Ron Soferman has launched a series of lectures to provide a robust yet simple overview of how to ensure that computer vision projects respect goals, budget and deadlines. Mr. Soferman shares one of his project management tips in each Computer Vision News issue. This month, we will learn about the Algorithmic Approach in Validation . One of the most interesting topics in computer vision validation is what we call the “ Algorithmic Approach ”. The idea behind it is the following: since the computer vision application is built out of blocks of algorithms and each one of them has its own assumptions, methods, strengths and weaknesses, we should consider the validation of the software as a task that is composed of the validation of each algorithm, following a step by step by step procedure. If we neglect the algorithmic steps and look at the module like a black box, we might be doomed to test and validate only the individual cases which we have met and not the other cases which could have been met, had a wider testing been performed. The Algorithmic Approach starts with understanding the algorithmic elements of each step in the module . This is done by taking the description of the algorithms, usually found in the ADD (Algorithmic Design Document) , which offers detailed documentation about every step and every algorithm, including all the parameters being used with them. ADD is also typically needed for compliance with standards like the FDA, CE and ISO, which require transparency about the software and the logic by which it is built. It is also a great tool for R&D management to follow the progress of the development during all its phases. Several actions are recommended to perform this Algorithmic Approach in validation. First, we use to do a peer review with algorithmic experts to assess each step, assumptions and weaknesses ad possibly errors to which the algorithm is exposed. It is done by playing the “ devil’s advocate ”, who does not take anything for granted. “ Consult with knowledge experts in the specific field, to understand all the possible scenarios that might be found ” 34 Computer Vision News Project Management Tip Validation: the Algorithmic Approach Management