NeRF-Loc: Transformer-Based Object Localization Within Neural Radiance Fields

TL;DR

NeRF-Loc introduces a transformer-based approach to accurately identify and localize objects within Neural Radiance Fields, enhancing robotic scene understanding for navigation and manipulation.

Contribution

The paper presents a novel transformer framework, NeRF-Loc, for extracting 3D object bounding boxes directly from NeRF models, improving over RGB(-D) based methods.

Findings

Outperforms baseline RGB(-D) methods in object localization accuracy.

Enables real-time training and object detection within NeRF scenes.

Uses dual transformer streams for context and detail encoding.

Abstract

Neural Radiance Fields (NeRFs) have become a widely-applied scene representation technique in recent years, showing advantages for robot navigation and manipulation tasks. To further advance the utility of NeRFs for robotics, we propose a transformer-based framework, NeRF-Loc, to extract 3D bounding boxes of objects in NeRF scenes. NeRF-Loc takes a pre-trained NeRF model and camera view as input and produces labeled, oriented 3D bounding boxes of objects as output. Using current NeRF training tools, a robot can train a NeRF environment model in real-time and, using our algorithm, identify 3D bounding boxes of objects of interest within the NeRF for downstream navigation or manipulation tasks. Concretely, we design a pair of paralleled transformer encoder branches, namely the coarse stream and the fine stream, to encode both the context and details of target objects. The encoded features are then fused together with attention layers to alleviate ambiguities for accurate object localization. We have compared our method with conventional RGB(-D) based methods that take rendered RGB images and depths from NeRFs as inputs. Our method is better than the baselines.