Google's libraries and pretrained models for application features such as pose landmarks, object detection, and audio classification. Task and platform support
Last reviewed: 2026-10-03
MediaPipe provides task libraries and pretrained models for application features such as pose landmarks and object detection. Pose Landmarker estimates body landmarks from supported inputs. Platform support, input modes, and customization options vary by task.
For a fitness prototype, begin with consented or synthetic test clips and a narrow movement. Compare landmark stability across camera angles, clothing, lighting, and occlusion. Measure latency on the intended device and avoid presenting pose estimates as a clinical or injury assessment.
No. Pose Landmarker estimates landmarks. Any coaching or health interpretation built on them requires its own evidence and evaluation.
No. Check the support table and input-mode documentation for the specific task and target platform.