This chapter addresses theLocalization core issue of localization in computer vision, which involves estimating the position of objects in an image correctly—not just classifying them. The chapter considers commonplace methods to get to localization, including bounding box regressionBounding box regression for coarse localization, instance segmentationInstance segmentation (masking) for pixel-level accuracy and landmark detection for the purpose of finding the specific location of object parts. The chapter discusses the role of confidence indices to filter the detections and critical quality ground truthGround truth data for training and evaluating computer vision models. The chapter further provides evaluation metricsEvaluation metrics that quantify localizationLocalization quality, such as the notion of Intersection over UnionIntersection over Union (IoU), and then the probabilistic foundations for models to quantify prediction uncertainty. The chapter discusses primary loss functionsLoss functions, including both Smooth L1 and IoUIntersection over Union based losses, that are iteratively optimized for bounding box estimation. The chapter covers a wide range of forces that underpin the critical idea of localizationLocalization so that the chapter serves to provide the foundational concepts needed to understand and then consider metrics useful for quantifying improvement in the accuracy of object localizationLocalization for computer vision in both modern and deep learningDeep learning systems.