Deep learning methods for feature extraction using computer vision are giving concrete results. This talk provides an overview of how feature detection from satellite imagery can provide a seamless experience for mappers in OpenStreetMap. Over the last year the Humanitarian OpenStreetMap Team facilitated an initiative to test using data derived using machine learning models to improve coordination of groups of mappers, quality of edits and overall experience when mapping in OpenStreetMap. During the talk the concrete results are shown and you will obtain a good idea about the state of machine learning data and tools around OpenStreetMap.