Introduction
The Rogue Mobile Vision SDK is a library that enables mobile applications on both iOS and Android devices to engage the camera to count reps and provide form-feedback to athletes.
It includes a registry of more than 100 movements that have been trained and validated with thousands of hours of real workouts, covering a wide range of athlete physiques, environmental conditions and exercise variations.
The library powered the Rogue Compete Challenge at the 2022 Crossfit Games where athletes compete head-to-head Thrusters and American Kettlebell Swings for time.
Use Cases
This library is intended to be integrated into partner's existing mobile applications. It can be used to facilitate a variety of use cases and applications:
- Automatic Workout Logging
- Workout Coaching Assistant
- Movement Form Improvement
- Competitive Judging
- Smart Mirrors
Design Goals
The library has been designed to operate on any modern smartphone. It does not require any additional wearable accessories or special input devices. It can work with both front and rear cameras.
Requirements
The Rogue Vision React Native SDK has dependencies on React Native, Core ML (iOS) and BlazePose which each have OS requirements of their own:
| Syntax | iOS | Android | Reference |
|---|---|---|---|
| React Native | 12.4+ | 5.0+ | requirements |
| MediaPipe Pose | Text | Android SDK 30.0 Android NDK 18-21 | requirements |
CoreML / iOS Versions: https://apple.github.io/coremltools/mlmodel/Format/Model.html#model
Validated Versions
The current version of the SDK is built and tested against:
| Platform | Version |
|---|---|
| iOS | 15-17 |
| Android | 6-15 |
| MLKit | 1.0 (iOS) / 18.0 (Android) |
| React Native | 0.64 |
| React Native Vision Camera | 2.9.x |
Limitations
Single Athlete
The library is designed to capture a single athlete at a time. If multiple athletes are in the frame during an exercise, it may cause errors in the output. It is designed to capture the closest person in frame.
Other Workout Equipment
The library does not support movements on mechanical workout equipment such as bikes, treadmills or rowers.
Walking and Running
The library does not support walking or running measurements.