Search:
Match:
2 results

Analysis

This paper addresses the challenge of reconstructing 3D models of spacecraft using 3D Gaussian Splatting (3DGS) from images captured in the dynamic lighting conditions of space. The key innovation is incorporating prior knowledge of the Sun's position to improve the photometric accuracy of the 3DGS model, which is crucial for downstream tasks like camera pose estimation during Rendezvous and Proximity Operations (RPO). This is a significant contribution because standard 3DGS methods often struggle with dynamic lighting, leading to inaccurate reconstructions and hindering tasks that rely on photometric consistency.
Reference

The paper proposes to incorporate the prior knowledge of the Sun's position...into the training pipeline for improved photometric quality of 3DGS rasterization.

Analysis

This paper introduces Local Rendezvous Hashing (LRH) as a novel approach to consistent hashing, addressing the limitations of existing ring-based schemes. It focuses on improving load balancing and minimizing churn in distributed systems. The key innovation is restricting the Highest Random Weight (HRW) selection to a cache-local window, which allows for efficient key lookups and reduces the impact of node failures. The paper's significance lies in its potential to improve the performance and stability of distributed systems by providing a more efficient and robust consistent hashing algorithm.
Reference

LRH reduces Max/Avg load from 1.2785 to 1.0947 and achieves 60.05 Mkeys/s, about 6.8x faster than multi-probe consistent hashing with 8 probes (8.80 Mkeys/s) while approaching its balance (Max/Avg 1.0697).