OBSER: Object-Based Sub-Environment Recognition for Zero-Shot Environmental Inference

TL;DR

OBSER is a Bayesian framework that enables zero-shot recognition of environments by inferring relationships between sub-environments and objects using learned object distributions, validated through theoretical and empirical methods.

Contribution

Introduces a novel Bayesian framework for sub-environment recognition that leverages object distributions for zero-shot environmental inference.

Findings

Outperforms scene-based methods in chained retrieval tasks

Reliable inference in open-world and photorealistic environments

Validates the ($psilon,elta$) statistically separable function

Abstract

We present the Object-Based Sub-Environment Recognition (OBSER) framework, a novel Bayesian framework that infers three fundamental relationships between sub-environments and their constituent objects. In the OBSER framework, metric and self-supervised learning models estimate the object distributions of sub-environments on the latent space to compute these measures. Both theoretically and empirically, we validate the proposed framework by introducing the ($\epsilon,\delta$) statistically separable (EDS) function which indicates the alignment of the representation. Our framework reliably performs inference in open-world and photorealistic environments and outperforms scene-based methods in chained retrieval tasks. The OBSER framework enables zero-shot recognition of environments to achieve autonomous environment understanding.