This dataset consists of a raster and surficial shapefiles for the Neversink River watershed, NY, that were generated using ArcGIS Pro's deep learning functionality. The shapefiles contain polygons that show the locations of shallow-to-exposed bedrock and alluvium-filled valleys, while the raster provides estimated minimum sediment thicknesses in the areas between the shallow/exposed bedrock and alluvium-filled valleys. The deep learning model that generated shallow/exposed bedrock and alluvium polygons was trained on existing geologic maps of New York and Pennsylvania. Minimum sediment thicknesses were estimated by applying an inverse distance weighting interpolation to measurements of stream incision depth and edges of shallow/exposed bedrock zones (where sediment thickness ≈ 0 m). In this area, stream incision through bedrock was considered negligible. The interpolation used variable numbers of points, a fixed interpolation distance of 250 m, and an output resolution of 30 m/pixel. Given the possibility that streams may have not completely incised through surficial materials down to bedrock, these estimates are minimum constraints on the true thicknesses of surficial sediments.