aiqbench analysis
FlatLands establishes a rigorous benchmark for generating complete metric floor maps from single egocentric images, directly addressing the limitations of partial field-of-view in embodied navigation systems. By comparing deterministic and stochastic generative models against training-free baselines, this work advances uncertainty-aware mapping techniques essential for robust indoor AI agents. The dataset's inclusion of visibility and validity metrics provides a critical testbed for evaluating how well current vision models handle occlusion and out-of-distribution scenarios in real-world environments.
FlatLands: Generative Floormap Completion From a Single Egocentric View
arXiv:2603.16016v3 Announce Type: replace-cross 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.
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.
Computer Vision and Pattern Recognition (cs.CV); Artificial Intelligence (cs.AI); Robotics (cs.RO); Image and Video Processing (eess.IV)
arXiv:2603.16016 [cs.CV]
(or arXiv:2603.16016v3 [cs.CV] for this version)
From: Subhransu S. Bhattacharjee Mr. [view email] [v1] Mon, 16 Mar 2026 23:50:16 UTC (7,095 KB) [v2] Sat, 27 Jun 2026 05:45:45 UTC (7,145 KB) [v3] Wed, 2 Sep 2026 09:21:19 UTC (7,145 KB)
Full text under the publisher's open licence · 174 words
Related coverage
MetaWorld: Scaling Multi-Agent Video World Model from Single-view Video Data
arXiv · Jun 3, 2026
Runway’s WorldPrompt and the Engineering of Real-Time Worlds
Latent Space · Sep 25, 2026
Gemini 4 Argon for Robotics
Google DeepMind · Oct 13, 2026
[AINews] Xiaomi MiMo-V2.6-Pro 1T-A42B: the new top Open Weights model, trained for $3M
Latent Space · Sep 22, 2026