We're in a transitional time where it is not always clear whether (Deep) Neural Networks are the best way to solve a given problem - even in Computer Vision. This talk discusses practical steps, tools and practices that can help you make that decision and follow through.
Engineers and researchers come in all shapes and colours, but we tend to have one thing in common: an eagerness to see nails for our new hammers everywhere. This talk discusses practical steps that can help you characterize your problem and decide whether to adopt a traditional Computer Vision approach or invest in a Deep Learning solution. It will then discuss tools, metrics and good practices in the implementation to validate reliably, even across very different technologies.
The talk is geared towards practitioners who start from a problem and wonder about the how. However, it can be useful to researchers and students who start from a technique they are studying and want some additional tools and engineering tricks to do so systematically.
It will assume some some familiarity with software development and with Machine Learning, Computer Vision and/or Deep Learning.