Computer Vision News - May 2019

12 Computer Vision News Project Management Tip Management When the development team delivers work to the client and/or to the testing team, it may happen that the software displays unwanted behavior or failure. We are talking about real-time software, a situation in which future input is unknown at the coding time. The development team should always keep in mind the possibility of a future crisis, which might occur later on. Sometimes, developers have the tendency to work on the urgent task, like delivery, functionality, communication with the customer and so on. This propension could be magnified when the software is performing fine with current data: all the cases are working OK and the customer (or the testing team) is very much satisfied with the performance of the software. When future input is unknown, the project manager should always be prepared to the possibility of a crisis when additional input data is provided. A good solution to never go unprepared is the accuracy measure: a measure which could define how large is the margin which keeps our algorithm on the safe side, with regards to future input data. These accuracy measures will give information beyond the current performance with current data: they will tell how close the software is to its maximal capability; in other words, how easy it is for the software to provide the current performance and how large are the margins until it might meet a crisis. A practical example for the accuracy measure: let’s say that we have an image and we are looking for a circle inside it. We want to measure how close is the found circle to our model of circle. Do all the points in this circle fit with the given formula of a circle? Once we have defined this accuracy measure, we need find a bulk of cases upon which test it and find out where we stand. We can add random changes to that bulk of results, so to include in the test also some levels of noise, that we expect to receive later in the real data. We increase the data complexity to check how far the software can still go. Ideally, the project manager should never find the software too close to the limits: there should always be a double margin with respect to the current results. This work on the stability of the software must be done, regardless of the high pressure being Crisis Management RSIP Vision’s 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 Nissim Avitan tells us about another aspect of successful Project Management: crisis management . It’s another handy tip by RSIP Vision for Project Management in Computer Vision . “…never go unprepared…”