1 Editorial The remarkable growth in computational power over the last decade has enabled important advances in machine learning, allowing us to achieve impressive results across all areas of image and video processing. Numerical methods that were once thought to be intractable are now commonly deployed to solve problems as diverse as 3D modeling from 2D data (photo tourism [1]), object recognition (logo detection [2]), human biometrics (face recognition [3]), and video surveillance (automatic threat detection [4]). Despite this tremendous progress, there are many open questions related to understanding visual data—we are still far from matching human visual ability in all of these areas. It is fair to scrutinize this observation in some detail— where are researchers and current methodologies falling short? From our perspective, the practicalities of realworld problems are often obscured by the mischaracterization of good results in limited contexts, theoretical frameworks built around artificial problems, and a deep sea of technical minutiae. Assumptions are a necessary component to problem solving, but poor ones lead our algorithms and subsequent analyses astray. Similarly, good theory is important, but we should not lose sight of the fundamental problemwe are trying to solve by abstracting it away. Such circumstances occur more often than one might think. How should a researcher charged with the design of an algorithm that must actually work outside of the laboratory avoid these dilemmas? Perhaps unsurprisingly, we, as academics, are typically more critical than constructive when evaluating newwork—this is especially true of paper reviews and survey articles. In response to this, we submit the following researcher’s guide to everyday machine learning as a gentle nudge forward: