ActiveInitSplat: How Active Image Selection Helps Gaussian Splatting

2025-03-10

AI Summary

This work addresses the limitation in Gaussian Splatting (GS) reconstruction quality arising from passive image selection during initialization. We propose the first active image selection framework tailored for GS initialization. Methodologically, we formulate scene representation quality as a feedback signal and integrate 3D density modeling with viewpoint coverage assessment to design an iterative selection algorithm. Images are dynamically chosen based on a density-occupancy criterion that jointly maximizes information content, viewpoint diversity, and structural alignment. To our knowledge, this is the first application of active learning to GS initialization, breaking away from conventional dense or random sampling paradigms. Extensive experiments on multiple synthetic and real-world datasets demonstrate consistent improvements over passive initialization baselines: LPIPS decreases by 18.7%, SSIM increases by 5.2%, and PSNR improves by 3.9 dB.

Technology Category

• Machine Learning: Active Learning
• Search and Optimization: Sampling/Simulation-based Search
• Computer Vision: Scene Analysis & Understanding

Application Category

• Search and Retrieval-Augmented AI: Web learning to rank, online learning, and counterfactual learning for ranking
• Graph Algorithms and Modeling for the Web: Efficient manipulation of static and dynamic Web-related graphs
• User Modeling, Personalization and Recommendation: Fairness-aware retrieval and ranking

Abstract

Gaussian splatting (GS) along with its extensions and variants provides outstanding performance in real-time scene rendering while meeting reduced storage demands and computational efficiency. While the selection of 2D images capturing the scene of interest is crucial for the proper initialization and training of GS, hence markedly affecting the rendering performance, prior works rely on passively and typically densely selected 2D images. In contrast, this paper proposes `ActiveInitSplat', a novel framework for active selection of training images for proper initialization and training of GS. ActiveInitSplat relies on density and occupancy criteria of the resultant 3D scene representation from the selected 2D images, to ensure that the latter are captured from diverse viewpoints leading to better scene coverage and that the initialized Gaussian functions are well aligned with the actual 3D structure. Numerical tests on well-known simulated and real environments demonstrate the merits of ActiveInitSplat resulting in significant GS rendering performance improvement over passive GS baselines, in the widely adopted LPIPS, SSIM, and PSNR metrics.

Problem

• Improves Gaussian splatting initialization via active image selection.
• Enhances 3D scene coverage using diverse viewpoint criteria.
• Boosts rendering performance metrics like LPIPS, SSIM, and PSNR.

Innovation

• Active selection of 2D training images
• Density and occupancy criteria for 3D representation
• Improved Gaussian splatting rendering performance