Oral Presentations and Performances: Session I
Project Type
Presentation
Project Funding and Affiliations
Montana Space Grant Consortium / NASA
Abstract / Artist's Statement
This report details the foundational development of a unified splatting ecosystem designed to enhance spatial mapping and 3D reconstruction accuracy for autonomous aerial systems. The primary objective was to bridge the gap between academic neural rendering algorithms and practical autonomous operations by building a decentralized, verifiable infrastructure. Methods included deploying containerized GPU workloads via the Akash network, integrating Solana-based on-chain data provenance, and shifting focus toward a photo/video-centric backend validation layer to ensure data sovereignty. Results indicate that the 3D rendering pipeline and core microservices architecture are functional in isolation, with the distributed network components successfully wired for future end-to-end testing. The project purposefully pivoted from immediate full-scale aerial deployment and 3D/4D model reconstruction to solidifying the hardware validation and provenance infrastructure, creating a sturdy foundation for user-facing applications. Future work will focus on integrating these validated backend processes into real-time image-to-3D/4D reconstruction pipelines.
Keywords: Convex Splatting, decentralized compute, data provenance, autonomous systems, spatial mapping
Category
Physical Sciences
Development of a Unified Splatting Ecosystem for Enhanced Spatial Mapping and Accuracy
UC 329
This report details the foundational development of a unified splatting ecosystem designed to enhance spatial mapping and 3D reconstruction accuracy for autonomous aerial systems. The primary objective was to bridge the gap between academic neural rendering algorithms and practical autonomous operations by building a decentralized, verifiable infrastructure. Methods included deploying containerized GPU workloads via the Akash network, integrating Solana-based on-chain data provenance, and shifting focus toward a photo/video-centric backend validation layer to ensure data sovereignty. Results indicate that the 3D rendering pipeline and core microservices architecture are functional in isolation, with the distributed network components successfully wired for future end-to-end testing. The project purposefully pivoted from immediate full-scale aerial deployment and 3D/4D model reconstruction to solidifying the hardware validation and provenance infrastructure, creating a sturdy foundation for user-facing applications. Future work will focus on integrating these validated backend processes into real-time image-to-3D/4D reconstruction pipelines.
Keywords: Convex Splatting, decentralized compute, data provenance, autonomous systems, spatial mapping