Single-Image Food Mass Estimators Do Not Transfer Their Gram Scale Across Datasets
Estimating food mass from a photograph is central to image-based dietary assessment, but whether trained models transfer to new cameras and foods is unclear.
ProofPaper ↗
Key points
- We fine-tuned DINOv2-base on 2,253 weighed Nutrition5k dishes to predict mass from one RGB photograph and compared it with a zero-shot vision-language model, Qwen3-VL-32B, on Nutrition5k and three external datasets: ASA24 studio portions, ECUSTFD whole foods and Food Portion Benchmark meal trays.
- DINOv2-large, a larger backbone trained on the same dishes, did not reduce the external error.
- One labeled photograph of the same food cut the model's ASA24 error to 28.3 g, better than that photograph achieved alone.
- Single-image mass estimators therefore need labeled photographs from the setting where they are used.
Sources (1)
- [1]Single-Image Food Mass Estimators Do Not Transfer Their Gram Scale Across DatasetsarXiv (AI, ML, NLP, CV, robotics, multi-agent) · Oct 21, 12:59 AM
Estimating food mass from a photograph is central to image-based dietary assessment, but whether trained models transfer to new cameras and foods is unclear.
We fine-tuned DINOv2-base on 2,253 weighed Nutrition5k dishes to predict mass from one RGB photograph and compared it with a zero-shot vision-language model, Qwen3-VL-32B, on Nutrition5k and three external datasets: ASA24 studio portions, ECUSTFD whole foods and Food Portion Benchmark meal trays.
Extractive summary: sentences quoted from the sources.
Before this
- Oct 20, 2026Near-Duplicate Image Selection Transfers Preferences in Vision-Language Models
- Oct 20, 2026You are the Team — Anekha Sokhal, Moshi
- Oct 20, 2026Qwen-Audio-3.1-ASR Technical Report
- Oct 19, 2026Structured outputs for Anthropic Claude models on Amazon Bedrock
- Oct 19, 2026Introducing Clairic, a new System One class of extraction models
- Oct 19, 2026MiniMax M3.1 Flash shows China's AI price race has no floor in sight