# Third-party notices

ATIANA Vision redistributes or downloads the following third-party components. The
public web demo serves the components marked **(public demo)** from its own origin.

| Component | Version / file | Licence | Basis | Source |
|---|---|---|---|---|
| MediaPipe Tasks Vision (JS + WebAssembly runtime) **(public demo)** | `@mediapipe/tasks-vision` 1.1.0 | Apache-2.0 | npm package licence field | https://www.npmjs.com/package/@mediapipe/tasks-vision |
| EfficientDet-Lite0 object detector (int8) **(public demo)** | `efficientdet_lite0.tflite` | Apache-2.0 | Distributed by Google in the MediaPipe model collection. Google publishes no model card or model-specific licence for it; the guide page licenses code samples under Apache-2.0 and page content under CC-BY-4.0. | https://developers.google.com/edge/mediapipe/solutions/vision/object_detector |
| MediaPipe Hand Landmarker **(public demo)** | `hand_landmarker.task` | Apache-2.0 | Model card, "Licensed under" | https://storage.googleapis.com/mediapipe-assets/Model%20Card%20Hand%20Tracking%20(Lite_Full)%20with%20Fairness%20Oct%202021.pdf |
| MediaPipe BlazeFace short-range **(public demo)** | `blaze_face_short_range.tflite` | Apache-2.0 | Model card, "Licensed under" | https://storage.googleapis.com/mediapipe-assets/MediaPipe%20BlazeFace%20Model%20Card%20(Short%20Range).pdf |
| MediaPipe Pose Landmarker lite (Python only) | `pose_landmarker_lite.task` | Apache-2.0 | Model card, "Licensed under" | https://storage.googleapis.com/mediapipe-assets/Model%20Card%20BlazePose%20GHUM%203D.pdf |
| SSD-MobileNet v1, opset 12 (Python only, optional) | `ssd_mobilenet_v1_12.onnx` | MIT | ONNX Model Zoo model page | https://github.com/onnx/models/tree/main/validated/vision/object_detection_segmentation/ssd-mobilenetv1 |

Copyright for the MediaPipe runtime and models: Copyright Google LLC. Licensed under
the Apache License, Version 2.0; the full text is in `licenses/Apache-2.0.txt`
(served at `/licenses/Apache-2.0.txt` in the public demo). These components are used
unmodified.

The object detectors were trained on COCO 2017 (annotations CC-BY-4.0). The model
cards for BlazeFace and BlazePose list surveillance and identity recognition as out
of scope; ATIANA Vision uses them for detection only.

Python dependencies (installed by pip, not redistributed here) include MediaPipe
(Apache-2.0), OpenCV (Apache-2.0), ONNX Runtime (MIT), NumPy (BSD-3-Clause), FastAPI
(MIT), Pydantic (MIT) and Uvicorn (BSD-3-Clause). Web dependencies include Next.js and
React (MIT).

SJA Robotics names and logos are trademarks of SJA Robotics and are not licensed
for reuse.