Reproduce “VolETA: One- and Few-shot Food 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
Ahmad AlMughrabi, Umair Haroon, Ricardo Marques, Petia Radeva
VolETA, developed by the Universitat de Barcelona and the Computer Vision Center, introduces a framework for estimating food volume from one or a few RGBD images. The system uses a reference object for scale determination and achieves a 10.97% Mean Absolute Percentage Error on the MTF dataset for its few-shot approach.
Key insights
- The few-shot approach provides accurate food volume estimates, achieving a 10.97% Mean Absolute Percentage Error (MAPE) on the MTF dataset.
- A two-step reference object scaling method, involving initial measurement and depth-based fine-tuning, reliably determines absolute scale for reconstructed food items.
- The keyframe selection process effectively reduces the number of processed frames by 60-80% without a notable impact on reconstruction quality, leading to computational savings.
Summary by alphaXiv.
Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.