The deep learning revolution has accelerated discovery in cell biology by allowing researchers to outsource their microscopy analyses to a new class of tools called cell segmentation models. The performance of these models, however, is often constrained by the limited availability of annotated data for them to train on. This limitation is a consequence of the time cost associated with annotating training data by hand. To address this bottleneck, we developed Cell-APP (cellular annotation and perception pipeline), a tool that automates the annotation of high-quality training data for transmitted-light (TL) cell segmentation. Cell-APP uses two inputs—paired TL and fluorescence images—and operates in two main steps. First, it extracts each cell’s location from the fluorescence images. Then, it provides these locations to the promptable deep learning model μSAM, which generates cell masks in the TL images. Users may also employ Cell-APP to classify each annotated cell; in this case, Cell-APP extracts user-specified, single-cell features from the fluorescence images, which can then be used for unsupervised classification. These annotations and optional classifications comprise training data for cell segmentation model development. Here, we provide a step-by-step protocol for using Cell-APP to annotate training data and train custom cell segmentation models. This protocol has been used to train deep learning models that simultaneously segment and assign cell-cycle labels to HeLa, U2OS, HT1080, and RPE-1 cells.