Back to Gap Map Infrastructure / Classic ML & computer vision

MediaPipe

open source / Overall score: 4.0(strong)

Google's framework and ready-made solutions for on-device machine learning on Android, iOS, the web and Python. Its Tasks APIs run pretrained models for face detection and landmarks, hand and pose landmarks, gesture recognition, holistic tracking, object detection, image classification, embedding and segmentation, interactive segmentation, audio and text classification and language detection, and it adds on-device LLM inference. Some solutions can be retrained with the Model Maker tool, which is no longer actively maintained.

Openness

5 medium confidence

5.0

license: Apache-2.0(OSI)
source: public(github.com/google-ai-edge/mediapipe)
core features withheld: no — no paid edition

Apache-2.0 and built from the public repository, which moved from the google account to google-ai-edge. Google sells no edition of it, and its solution model cards, such as Hand Tracking's, carry Apache-2.0.

https://raw.githubusercontent.com/google-ai-edge/mediapipe/master/LICENSE recorded 2026-09-27
LICENSE is the Apache License, Version 2.0, with an appended notice for bundled Lucent Technologies utf code.
https://raw.githubusercontent.com/google-ai-edge/mediapipe/master/README.md recorded 2026-09-27
README: "MediaPipe Solutions provides a suite of libraries and tools for you to quickly apply artificial intelligence (AI) and machine learning (ML) techniques in your applications"; "MediaPipe models: Pre-trained, ready-to-run models for use with each solution"; no paid tier mentioned.
https://repos.ecosyste.ms/api/v1/hosts/GitHub/repositories/google-ai-edge/mediapipe recorded 2026-09-27
Repository record for google-ai-edge/mediapipe, not archived and not a fork, license apache-2.0, last pushed 2026-09-25.
https://storage.googleapis.com/mediapipe-assets/Model%20Card%20Hand%20Tracking%20(Lite_Full)%20with%20Fairness%20Oct%202021.pdf recorded 2026-09-27
Hand Tracking model card: "LICENSED UNDER Apache License, Version 2.0".
https://ungh.cc/repos/google-ai-edge/mediapipe/files/master recorded 2026-09-27
Repository tree of 4,866 paths with no ee, enterprise, commercial or pro directory.

Adoption

4 high confidence

4.0

Measured on monthly PyPI downloads of the mediapipe package. Android, iOS and web installs, which are most of MediaPipe's deployments, are not counted, so this is a floor.

https://pypistats.org/api/packages/mediapipe/recent recorded 2026-09-27
last_month 1,838,759 downloads of mediapipe

Capability

4 medium confidence

4.0

MediaPipe is built to be dropped into an app and used as is. Its landmark trackers and its interactive segmenter, which cuts out whatever object the user taps, work on any image with no training, which is what places it beside Ultralytics. Its object detector and classifier predict fixed benchmark classes, and retraining them relies on Model Maker, which Google no longer actively maintains.

https://ai.google.dev/edge/mediapipe/solutions/guide recorded 2026-09-27
Solutions guide: vision tasks face detection, face landmark detection, gesture recognition, hand landmark detection, holistic landmark detection, image classification, image embedding, image segmentation, interactive segmentation, object detection and pose landmark detection; text and audio tasks; generative AI tasks including LLM Inference.
https://ai.google.dev/edge/mediapipe/solutions/model_maker recorded 2026-09-27
Model Maker page: "Deprecated: MediaPipe Model Maker is still available, but is no longer actively maintained."

Verified 2026-09-27