FlatLands: Generative Floormap Completion From a Single Egocentric View

TL;DR

FlatLands introduces a comprehensive dataset and benchmark for completing full indoor floor maps from a single egocentric view, enabling better indoor navigation and mapping applications.

Contribution

The paper presents a new dataset, benchmark, and an end-to-end RGB-to-floor map pipeline for single-view indoor floor completion tasks.

Findings

Comparison of various approaches including training-free, deterministic, ensemble, and generative models.

FlatLands dataset contains over 270,000 observations from diverse indoor scenes.

Benchmark facilitates evaluation of uncertainty-aware indoor mapping methods.

Abstract

A single egocentric image typically captures only a small portion of the floor, yet a complete metric traversability map of the surroundings would better serve applications such as indoor navigation. We introduce FlatLands, a dataset and benchmark for single-view bird's-eye view (BEV) floor completion. The dataset contains 270,575 observations from 17,656 real metric indoor scenes drawn from six existing datasets, with aligned observation, visibility, validity, and ground-truth BEV maps, and the benchmark includes both in- and out-of-distribution evaluation protocols. We compare training-free approaches, deterministic models, ensembles, and stochastic generative models. Finally, we instantiate the task as an end-to-end monocular RGB-to-floormaps pipeline. FlatLands provides a rigorous testbed for uncertainty-aware indoor mapping and generative completion for embodied navigation.