# Third-Party Notices and Provenance

This file records the principal upstream software, model, and dataset sources used to create the released checkpoint. It is not legal advice.

## Ultralytics YOLO11

- Base checkpoint: `yolo11n-seg.pt`
- Training/export framework: Ultralytics 8.3.0
- Source and licensing information: https://github.com/ultralytics/ultralytics and https://www.ultralytics.com/license

Ultralytics states that models trained with its software are AGPL-3.0 by default unless covered by an Enterprise License. For that reason, this model repository uses GNU AGPL-3.0. This notice does not determine the licensing obligations of any larger application that consumes the model; downstream users must review their own distribution and deployment arrangement.

## FoodSeg103

- Dataset card: https://huggingface.co/datasets/EduardoPacheco/FoodSeg103
- License shown on that dataset repository: Apache-2.0
- Use in this run: all non-background semantic regions were merged into the single `measurable_object` foreground class.

No FoodSeg103 source images or annotations are redistributed here.

## COCO 2017

- Project site: https://cocodataset.org/
- COCO API: https://github.com/cocodataset/cocoapi
- Use in this run: selected food-positive categories and container-focused hard negatives.

The COCO API software license and the licenses attached to individual images are separate matters. No COCO source images or annotations are redistributed here.

## Released artifacts

The repository contains only derived model weights/Core ML export, training code, an executed notebook, run metadata, metrics, and documentation. It deliberately excludes raw dataset images, full dataset annotation copies, and personal test photographs.

GNU AGPL-3.0 for this repository does not replace or waive third-party licenses, terms, attribution requirements, privacy/publicity rights, or trademarks. No warranty of non-infringement is provided.