Reproduce “Denoising-Aware Temporal Point Cloud Completion for 3D Crop Architecture Recovery and Phenotypic Trait Extraction” on OpenResearch
Turn your favorite coding agents into research agents that review literature, develop hypotheses, and run experiments.
Universal app for Apple silicon and Intel. Requires macOS 11 or later.
About this paper
Mrudul Mittal, Soumyashree Kar
Researchers introduce a two-stage pipeline integrating denoising and temporal point cloud completion, specifically designed for agricultural crop architecture recovery from noisy, occluded sensor data. This approach, evaluated using a new synthetic dataset and real-world scans, leads to substantial improvements in the accuracy of critical phenotypic trait extraction.
Key insights
- Integrating a denoising stage substantially improves point cloud reconstruction quality, reducing real-data Chamfer Distance by up to 63.3%.
- Leveraging temporal information from prior growth stages through cross-attention in the **Adaptive Temporal PoinTr** network consistently enhances completion accuracy.
- The optimal denoiser choice is dataset-dependent, with GCN performing best on real-world structured sensor artifacts and **Mamba-DG** on synthetic isotropic noise.
Summary by alphaXiv.
Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.