Images, which are captured by an image capture device on a ground disturbing work machine, and that are taken from different perspectives, and are received. A machine learned image identification model identifies items of interest in the images. A three-dimensional representation is generated based on the set of images. The three-dimensional representation identifies a depth at which the recognized items lie beneath the surface of the soil being excavated. A map request is received which identifies a location and depth for which image data is to be provided. A three-dimensional representation of the location and depth are provided in response to the request.