Food portion segmentation: Datasets & Models
Instance segmentation of individual food items on a plate or tray, outlining each portion so its area and boundaries can be measured. Useful for automating portion control, nutrition estimation, and meal auditing in food service and packaging.
Food portion segmentation
This segmentation workflow generates a pixel mask for each individual food portion in an image, isolating items such as grains, proteins, and vegetables on a plate or tray. It accepts overhead or angled photos of plated meals, cafeteria trays, or packaged food, and returns per-item masks whose area and boundaries can be measured, making it a practical starting point for building on a food portion dataset.
Try it out
Powered by general detection model
Run on custom image
Drop an image here or click to upload
Why automate food portion segmentation
Manually checking that every plate meets a portion spec is slow, subjective, and impossible to do consistently at scale across a busy kitchen or production line. Trained inspectors disagree on what a 'standard scoop' looks like, and no one can eyeball grams of a side dish reliably meal after meal. Automating this with instance segmentation gives you a repeatable, quantitative measurement of each portion directly from an image, so under- and over-portioning are caught the moment a plate is assembled. For food and beverage operations, that means tighter cost control on ingredients and defensible nutrition and allergen data.
Key benefits
Pixel-accurate masks let you estimate portion area and relative volume rather than just detecting that food is present, and the model integrates with the overhead or line cameras you likely already run. Because it is segmentation-based, it handles crowded plates and touching items where box detectors fail, and it scales from a single prep station to every line in a facility without adding staff. Start from an existing food portion dataset on Roboflow Universe and fine-tune RF-DETR to your specific menu and camera setup.
How to use it
- Upload an image: Drop any photo into the widget above to try food portion segmentation for free, right on this page.
- Preview instant results: The model detects rice portion, chicken breast, vegetable serving, french fries, salad within seconds and returns an annotated image with structured predictions, so you can immediately see how well it works. Adjust the detection classes and re-run until the results fit your scenario.
- Like what you see?: With a free Roboflow account, you can deploy this workflow with the Serverless Cloud API, customize it in your workspace, or train RF-DETR (instance segmentation) on your own images for production accuracy.
- Get started for free: No code or training data required to begin: upload, review the results, and make it part of your stack in a single click.
Recommended model for food portion segmentation
RF-DETR is state of the art for real-time object detection, instance segmentation, and keypoint detection, with higher accuracy and lower latency than comparable models, and it's Apache 2.0 licensed. When you're ready to build a production model for food portion segmentation, fine-tune it on your own labeled images, then deploy to a Serverless Cloud API endpoint.
Not sure where to start?
Describe your Food Portion Segmentation project to Roboflow Agent.