InObject detection models this chapter, we explore algorithms used in object detection pipelines, both single-stage and two-stage architectures. We begin with an overview of single-stage detectors such as Single Shot Multibox DetectorSingle shot multibox detector (SSD) and EfficientDetEfficientDet, as they employ a single pass for localizationLocalization and classificationClassification processes to prioritize fast inference, as the ideal application of edge devicesEdge devices for real-time applications. After reviewing single-stage architectures, we then focus on two-stage architecture, specifically, the R-CNNR-CNN family of detectors, which take precedence over accuracy by creating region proposals to localize objects before classifying each object, and at the price of inference time. We will then cover some specific architectural components that lead towards and advance the two-stage architecture, such as region proposal networksRegion proposal network (RPNs), as significant to FasterR-CNN R-CNNFaster R-CNN, and feature pyramid networksFeature pyramid networks (FPNs), as part of MaskR-CNN R-CNNMask R-CNN, to demonstrate improvements in various tasks such as multi-scale object detection and instance segmentationInstance segmentation. Finally, we will only briefly mention anchor-point free models, such as CenterNetCenterNet, that anticipate that in most cases, one point can represent that object, thereby stripping down detection pipelines. The paper examines how speed, accuracy, and computation efficiency involve good trade-offs, and highlights where certain models are better suited for some application, e.g. autonomous drivingAutonomous driving vs medical imagingMedical imaging. At the end of the paper, the authors review deployment frameworksFrameworks like TensorFlowTensorFlow LiteTensorFlow Lite, which allow the optimizationOptimization and use of these complex models on lower resource production hardware, thus allowing a means of getting academia and technological investment into usable/practical solutions for the real-world.