DeepREM: Deep-Learning-Based Radio Environment Map Estimation from Sparse Measurements DeepREM combines two deep-learning models (U-Net and CGAN) that estimate Radio Environment Maps (REMs) from sparse measurements. In this repository, we present the dataset and the interactive app developed to use the resulting models derived from the research. Urban REMs dataset: We present a (REM) dataset of urban scenarios. Each map provides coverage information in areas from 2290 x 3670 m2 to 3810 x 5160 m2 with a resolution of 10 m. Coverage areas are sampled from Colombian cities (Armenia, Bogota, Cali, Ibague, Manizales, Medellin, and Pasto) and U.S. cities (Columbus and Washington). The simulations include topographic and building vector database information and Intelligent Ray-Tracing as a propagation model. In our second version, 400 new REMs were added to the dataset, including 4 new city areas with different topographic and building features (100 REMs for each area). The new cities are Barranquilla, Bucaramanga, Popayan, and North Pasto, all in Colombia. In addition, we also present an interactive application developed in the Streamlit framework to test CGAN and UNET performance in RSRP and BS coverage predictions either with REMs from our dataset or with completely new user-supplied REMs.The repository of the app can be downloaded at the following link: DeepREMapp