mbforbes / mapgen

Procedural map generation with GANs.
https://maxwellforbes.com/posts/procedural-map-generation-with-gans
13 stars 0 forks source link

mapgen

Procedural map generation with GANs.

demo

Code is in src/, I've rendered the report/ for easier reading on my website.

relwork

General

Modeling the Appearance and Behavior of Urban Spaces (Vanegas et al., 2009) [pdf]

Overview. Breaks problems into "Layout Modeling" (road networks + land use labels) (my focus), "Building Modeling" (3D geometry of buildings), and "Facade Modeling" (3D facades + textures).

Layout Modeling

Interactive Geometric Simulation of 4D Cities (Weber et al., 2009) [pdf]

Simulation of a 3D city growing over time.

Purdue group w/ several relevant papers

Some more details are below.

Example-Driven Procedural Urban Roads (Nishida et al., 2015) [pdf]

Tool for editing road networks.

Procedural Generation of Parcels in Urban Modeling (Vanegas et al., 2012) [pdf]

Algorithms for automatically splitting blocks (spaces inbetween a road network) into parcels.

Interactive Example-Based Urban Layout Synthesis (Aliaga et al., 2008)

Given an example image, synthesize a street network and copy and paste aerial pictures into them.

Interactive Reconfiguration of Urban Layouts (Aliaga et al., 2008)

A tool to made edits to an urban layout and regenerate / conform the roads / parcels / images to look nice again.

Learning

A Probabilistic Model for Exteriors of Residential Buildings (Fan and Wonka, 2016) [pdf] [prj]

Learn a GM for 3D buildings and facades.

Inverse Procedural Modeling of Facade Layouts (Wu et al., 2014) [pdf]

Learn (I think) a split grammar for facades, preferring shorter descriptions.

Bayesian Grammar Learning for Inverse Procedural Modeling (Martinovic and Gool, 2013) [pdf]

Learn a 2D CFG from a set of labeled facades.

A Probabilistic Model for Component-Based Shape Synthesis (Kalogerakis et al., 2012) [pdf]

Learn GM on shapes and generate their own (shapes like dinosaur, chair, boat).

Inverse Procedural Modeling by Automatic Generation of L-systems (Stava et al., 2010) [pdf]

(Not actually learning, I think) Constructs an L-system from a 2D vector image.

Detecting and Parsing Architecture at City Scale from Range Data (Toshev et al., 2010) [pdf]

Learn to parse point clouds to tree of roofs.