Visual and spatial data, such as images, photographs and maps, are playing important roles in digital humanities to understand the culture and landscape in the past. In comparison to textual data, where machines are actively used for reading in the form of text mining and “distant reading,” visual and spatial data are now dependent on human reading due to the lack of technology that helps human interpretation. In this talk, I will introduce machine-assisted methods to explore a new way of reading visual and spatial data, such as IIIF (International Image Interoperability Framework) for curating data, machine learning for extracting information, and digitally-enabled criticism to evaluate the content. I will finally introduce our ongoing project, called “historical big data,” to reconstruct the past such as the Edo Period in Japan (1603-1868) in terms of city, people, climate and disasters.