Computer Vision News - December 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. This month we learn about the Proof of Concept . The Proof of Concept (POC) is a feasibility test. The subject is interesting and useful because in the life cycle of R&D activity we always try to check whether there is any additional technology which could be used to solve problems or provide new features to our software. I could not stress enough the importance of starting a POC with very clear goals, definite timetable and a precise work plan. I mention this subject because it also belongs to the realm of research, with characteristics of open-end investigation. It is true that when you start to dig into a problem, you don’t know yet how you are going to solve it, but if you plan it properly and your requirements are well defined , you are in better position to progress efficiently. What are we going to prove with a POC? Which cases are we going to work on during the phases of the POC? We are not solving anything yet, but it is crucial to define the exact data and the exact device that we are going to work on, as well as the exact pathologies or phenomena that we want to detect. The availability status of needed data and the ground truth with which we will perform the validation process must be very clear at this stage. Timetable is key: when we estimate the resources and the feasibility of solutions , we have to make sure that we meet deadlines even at the level of initial research. The developer’s strong beliefs regarding his or her next steps might hamper the feasibility, most developers underestimating the time needed to reach a full solution. Time definitions must therefore be set in advance. If some specific results have already been obtained and we believe that they are meaningful, we can think of modifying the timeline. But if no essential results to lean on are at hand, we might suspect that this proof of concept is not working well. In that case, the options are to modify part of the definitions or to switch to a new plan altogether . Sometimes, timelines are influenced by technical problems. Especially when data to be used during research is not available yet. Whatever was the level of confidence in collecting the images, it might happen that the device is not well-prepared, that privacy issues arise or simply that images are not good enough. In this case, it is better to freeze the research temporarily, rather than investing too much effort before the infrastructure is ready . My recommendation to project managers is to plan the POC very carefully : by that, I believe that they will get the best results . “Starting a POC with very clear goals, definite timetable and a precise work plan” 25 Computer Vision News Project Management Tip POC - Proof of Concept Management