ML Kit Pose Detection Makes Staying Active at Home Easier

The ML Kit Pose Detection API is a new feature added to Google's ML Kit, designed to facilitate the integration of on-device machine learning into mobile apps, enhancing user experiences by tracking physical actions in real-time. Compatible with both Android and iOS, this lightweight solution offers a 33-point skeletal match that includes facial landmarks and tracks hands and feet, making it ideal for applications such as fitness and yoga trackers. Utilizing the BlazePose pipeline, the API offers two operating modes—"Fast" for high frame rates and "Accurate" for stable coordinates—and supports both video and static image use cases. It optimizes performance through a two-step detection process, combining face detection with a prominent person detection algorithm, and includes a streaming mode for further latency improvements. Early adopters like Groovetime and adidas Training are using the API to provide real-time feedback on dance moves and virtual workouts, respectively, highlighting its potential in various real-world applications.