GO-PRE: Goal-Oriented Next-Best-View Selection via Predictive Rendering Entropy for Active 3D Reconstruction

2026-07-31

AI Summary

Existing active 3D reconstruction methods rely on proxy signals for viewpoint selection that are misaligned with the ultimate goal of rendering quality, hindering optimal reconstruction fidelity. This work proposes a goal-oriented next-best-view framework that, for the first time, employs predicted rendering entropy as a measure of information gain. By directly maximizing the reduction in predicted rendering entropy over a user-specified target view manifold, the method optimizes information acquisition in prediction space. It supports interactive goal specification and integrates average marginal predictive entropy reduction, target manifold modeling, and real-time information gain computation to enable efficient viewpoint planning. Experiments demonstrate that the approach significantly outperforms state-of-the-art methods across multiple benchmarks, achieving higher reconstruction accuracy and more reliable uncertainty quantification.

Abstract

Active 3D reconstruction relies on active view selection to maximize reconstruction fidelity under limited capture budgets. However, most existing methods rely on surrogate signals such as parameter uncertainty or geometric heuristics, but these signals are often misaligned with the ultimate goal: the fidelity of rendered predictions. We propose GO-PRE, a goal-oriented next-best-view selection framework that explicitly targets information gain in the prediction space. Specifically, we formulate the objective as maximizing the reduction of the average marginal predictive entropy over a user-specified target view manifold. GO-PRE supports interactive goal specification and yields an efficient acquisition rule that enables real-time computation of information gain. Extensive experiments across benchmarks demonstrate that GO-PRE consistently improves active reconstruction performance and provides more reliable uncertainty quantification compared to state-of-the-art methods.

Problem

• active 3D reconstruction
• next-best-view selection
• rendering fidelity
• predictive entropy
• goal-oriented

Innovation

• goal-oriented
• predictive entropy
• next-best-view
• active 3D reconstruction
• information gain