Training and evaluating spatial reasoning in embodied agents requires diverse environments that are both geometrically faithful and semantically queryable. Synthetic simulators offer ground truth semantics but sacrifice realism; simulators based on reconstructions of real-world environments have realistic appearance but lack ground truth semantics by default. We propose using Semantic Radiance Fields (SRF) as simulators for spatial reasoning agents. SRFs are a representation that unifies these requirements by lifting 2D semantic segmentations from pretrained vision models into a 3D radiance field that jointly encodes geometry, appearance, and per-class semantic identity. The resulting fields are reconstructed from posed RGB captures of real scenes and support novel-view synthesis, semantic and free-space queries within a single grounded representation. This enables the efficient generation of diverse real-world environments to train and evaluate spatial reasoning models. As an example application, we outline an SRF-driven simulator for an orchard apple-reaching task, in which the radiance field supplies camera rendering, semantic ground truth, and occupancy queries to a physics engine.
Nico Heider, Michal Jan Wlodarczyk, Katarzyna Wasielewska-Michniewska et al.· 0 citations
In recent years, further rapid growth of the number of available data sources has been observed. Among them, solutions based on the Internet of Things (IoT) start to play an increasingly important role, forming the basis of the Cloud-Edge-IoT continuum. Because of the large volumes and wide variety of data that IoT devices offer, efficient and flexible methods must be developed for data sharing, processing, and analysis (both within and outside of IoT). In this context, the concept of the architecture of the Federated Data Fabric, which focuses on manageability and flexibility of data processing, enabling distribution and federation of data processing components across the Cloud-Edge-IoT Continuum is presented. It provides all essential components, including (1) data catalog, (2) data-as-a-product management, and (3) data processing pipelines. Furthermore, the proposed architecture is capable of seamlessly combining heterogeneous data models and efficiently handling large volumes of batch and streaming data by offering efficient semantic annotation and translation capabilities. The proposed architecture is presented in the context of the EU-funded projects aerOS, where it was first implemented and empirically validated, and O-CEI, where it is currently under active development.
W. Pawłowski, Paweł Szmeja, Ignacio Lacalle et al.· Applied Sciences· 0 citations
We use cookies to run the site and, with your consent, for analytics and to show ads.
See our Cookie Policy.
