Abstract
We introduce ResPlan, a large-scale dataset of 17,000 detailed, structurallyrich, and realistic residential floor plans, created to advance spatial AIresearch. Each plan includes precise annotations of architectural elements(walls, doors, windows, balconies) and functional spaces (such as kitchens,bedrooms, and bathrooms). ResPlan addresses key limitations of existingdatasets such as RPLAN (Wu et al., 2019) and MSD (van Engelenburg et al., 2024)by offering enhanced visual fidelity and greater structural diversity,reflecting realistic and non-idealized residential layouts. Designed as aversatile, general-purpose resource, ResPlan supports a wide range ofapplications including robotics, reinforcement learning, generative AI, virtualand augmented reality, simulations, and game development. Plans are provided inboth geometric and graph-based formats, enabling direct integration intosimulation engines and fast 3D conversion. A key contribution is an open-sourcepipeline for geometry cleaning, alignment, and annotation refinement.Additionally, ResPlan includes structured representations of room connectivity,supporting graph-based spatial reasoning tasks. Finally, we present comparativeanalyses with existing benchmarks and outline several open benchmark tasksenabled by ResPlan. Ultimately, ResPlan offers a significant advance in scale,realism, and usability, providing a robust foundation for developing andbenchmarking next-generation spatial intelligence systems.