Due to the omnipresence of Neural Radiance Fields (NeRFs), the interest towards editable implicit 3D repre-sentations has surged over the last years. However, editing implicit or hybrid representations as used for NeRFs is difficult due to the entanglement of appearance and geom-etry encoded in the model parameters. Despite these chal-lenges, recent research has shown first promising steps to-wards photorealistic and non-photorealistic appearance ed-its. The main open issues of related work include limited in-teractivity, a lack of support for local edits and large memory requirements, rendering them less useful in practice. We address these limitations with LAENeRF, a unified framework for photorealistic and non-photorealistic appearance editing of NeRFs. To tackle local editing, we leverage a voxel grid as starting point for region selection. We learn a mapping from expected ray terminations to final output color, which can optionally be supervised by a style loss, resulting in a framework which can perform photorealistic and non-photorealistic appearance editing of selected re-gions. Relying on a single point per ray for our mapping, we limit memory requirements and enable fast optimization. To guarantee interactivity, we compose the output color using a set of learned, modifiable base colors, composed with additive layer mixing. selection. Compared to concurrent work, LAENeRF enables recoloring and stylization while keeping processing time low. Furthermore, we demonstrate that our approach surpasses baseline methods both quanti-tatively and qualitatively.

Abstract The recent research explosion around implicit neural representations, such as NeRF, shows that there is immense potential for implicitly storing high‐quality scene and lighting information in compact neural networks. However, one major limitation preventing the use of NeRF in real‐time rendering applications is the prohibitive computational cost of excessive network evaluations along each view ray, requiring dozens of petaFLOPS. In this work, we bring compact neural representations closer to practical rendering of synthetic content in real‐time applications, such as games and virtual reality. We show that the number of samples required for each view ray can be significantly reduced when samples are placed around surfaces in the scene without compromising image quality. To this end, we propose a depth oracle network that predicts ray sample locations for each view ray with a single network evaluation. We show that using a classification network around logarithmically discretized and spherically warped depth values is essential to encode surface locations rather than directly estimating depth. The combination of these techniques leads to DONeRF, our compact dual network design with a depth oracle network as its first step and a locally sampled shading network for ray accumulation. With DONeRF, we reduce the inference costs by up to 48× compared to NeRF when conditioning on available ground truth depth information. Compared to concurrent acceleration methods for raymarching‐based neural representations, DONeRF does not require additional memory for explicit caching or acceleration structures, and can render interactively (20 frames per second) on a single GPU.

This paper introduces an approach to generating environmental maps based on ultrasonic range data. By means of a learning classifier ultrasonic range data are condensed yielding abstract concepts which enable a mobile robot to discern situations. As a consequence the free-space can be partitioned into situation areas which are defined as regions wherein a specific situation can be recognized. Using dead-reckoning such situation areas can be attached to graph nodes generating a map of the free-space in the form of a graph representation. How the extended Kalman filter algorithm can be applied in this context to compensate the dead-reckoning drift is also discussed.