Maryam Hosseini

Open source computational tools and critical methods for understanding cities through imagery, geospatial data, and planning evidence.

Cities are not equally visible in planning data. Planning often depends on evidence produced through uneven abstractions: some infrastructures are standardized, counted, and modeled with great precision, while others enter the record unevenly, if at all.

We build open source computational tools that turn visual and spatial records of cities into planning evidence. Much of this work uses computer vision and representation learning, but the deeper question is what happens in the abstraction: who becomes visible, who is masked, and when a model's apparent pattern is only a spurious correlation with real planning costs. My work asks when these tools should be used, when they stop producing useful knowledge, and what standard of evidence they must meet before shaping planning decisions.