jimmie

## AI-Powered Upper-Body Muscle Annotation Understanding human muscle anatomy can be challenging, especially for fitness enthusiasts who want to identify the specific muscle groups they are training. Our project uses **Meta's Muse Spark 1.3 and MediaPipe** to transform ordinary photographs into interactive anatomical muscle maps. Users simply upload two photographs of their upper body, one from the front and one from the back. Muse Spark analyzes the images to identify and locate muscle groups, while MediaPipe detects body landmarks to guide anatomical positioning. Our custom annotation pipeline combines these outputs to generate color-coded muscle overlays directly on the user's photographs. Users can interact with individual muscle regions to learn their anatomical names and locations, including the pectoralis major, deltoids, biceps, triceps, trapezius, latissimus dorsi, and abdominal muscles. The system also distinguishes visible surface anatomy from deeper muscles whose positions can only be estimated. **Our goal is to make muscle anatomy more accessible, visual, and personalized through AI.** Rather than studying generic anatomical diagrams, users can explore muscle groups directly on their own bodies. We are also experimenting with Muse Spark's image-grounding capabilities to improve muscle localization and develop more anatomically accurate annotations. Future extensions include exercise recommendations tailored to selected muscle groups, helping users connect anatomical understanding with practical fitness training. **Built with:** Meta Muse Spark 1.3, MediaPipe, Python, FastAPI, Next.js, and TypeScript.