Reproduce “VolE: A Point-cloud Framework for Food 3D Reconstruction and Volume Estimation” 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
Umair Haroon, Ahmad AlMughrabi, Thanasis Zoumpekas, Ricardo Marques, Petia Radeva
The VolE framework enables reference-free and depth-free food volume estimation using widely available AR-capable mobile devices by combining structure from motion with specialized food segmentation. It achieves a Mean Absolute Percentage Error (MAPE) of 2.22% on the MetaFood dataset, outperforming existing approaches.
Key insights
- Combining AR-provided camera pose with Structure from Motion refinement allows for accurate 3D reconstruction at a real-world scale.
- Precise food video segmentation is essential for isolating the food object, significantly improving the accuracy of 3D reconstruction by removing background noise.
- The volume of the reconstructed food item can be accurately calculated from a watertight 3D mesh using the divergence theorem, enabling a reference-free and depth-free approach.
Summary by alphaXiv.
Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.