About
DanceVision is an AI-driven dance proficiency assessment system that leverages MediaPipe's pose estimation framework alongside various machine learning models to analyze and score dance performance. Built as an open-source Python project, it processes video input to extract skeletal pose landmarks from a dancer's body in real time or from recorded footage. The system works by training on reference dance videos — such as K-Pop choreography — and then predicting how closely a user's movements match the target performance. It includes scripts for data collection, model training, pose prediction demos, and video annotation, making it a complete end-to-end pipeline for dance analysis. Key capabilities include frame-by-frame pose extraction, ML model training on custom dance datasets, proficiency scoring, and annotated video output for visual feedback. The included Jupyter Notebook provides a comprehensive walkthrough of the full pipeline. DanceVision is ideal for developers and researchers exploring human motion analysis, fitness app builders, dance educators seeking automated feedback tools, and computer vision enthusiasts interested in applying pose estimation to creative domains. Being fully open source, it can be extended to support additional dance styles, scoring metrics, or deployment into interactive applications.
Key Features
- MediaPipe Pose Estimation: Extracts 33 body landmark points per frame using Google's MediaPipe framework for accurate skeletal tracking.
- ML-Based Proficiency Scoring: Trains machine learning models on reference dance videos to predict and score how closely a dancer's movements match the target choreography.
- Annotated Video Output: Generates annotated videos overlaying pose landmarks and assessment results for clear visual feedback on performance.
- End-to-End Pipeline: Includes scripts for training data collection, model training, pose prediction demos, and dance recording in a fully modular workflow.
- Jupyter Notebook Walkthrough: A comprehensive notebook guides users through the entire pipeline from data preparation to model evaluation.
Use Cases
- Automating dance skill assessment for K-Pop, fitness, or performing arts education platforms
- Researchers studying human motion analysis and pose estimation applied to expressive movement
- Developers building fitness or dance training apps that require pose-based feedback
- Computer vision students learning end-to-end ML pipelines using real-world video data
- Dance instructors exploring AI tools to provide objective, data-driven feedback to students
Pros
- Fully Open Source: Free to use, modify, and extend under an open-source license, making it accessible to developers and researchers.
- Complete ML Pipeline: Covers everything from data collection and model training to inference and video annotation, reducing setup friction.
- Built on Proven Libraries: Leverages MediaPipe, a production-grade pose estimation framework, ensuring reliable and accurate landmark detection.
Cons
- No GUI or Web Interface: Requires Python scripting knowledge to run; there is no ready-made user interface for non-technical users.
- Limited Dance Style Coverage: Currently demonstrated primarily on K-Pop choreography; extending to other dance styles requires custom data collection and retraining.
- Early-Stage Project: With only 5 commits and a small contributor base, the project may lack robustness and ongoing maintenance.
Frequently Asked Questions
What technology does DanceVision use for pose detection?
DanceVision uses Google's MediaPipe library to extract 33 skeletal body landmarks per video frame, which are then fed into machine learning models for proficiency assessment.
Can I train DanceVision on my own dance style?
Yes. The project includes data collection and model training scripts, allowing you to record reference videos of any dance style and train a custom model for proficiency scoring.
What programming language and environment does it require?
DanceVision is built in Python. Users need a Python environment with libraries such as MediaPipe, along with Jupyter Notebook for the guided walkthrough.
Is DanceVision suitable for real-time analysis?
The architecture supports real-time pose prediction via the included demo script, though real-time performance depends on your hardware capabilities.
Is this project free to use?
Yes, DanceVision is fully open source and hosted on GitHub, free to use, fork, and adapt for your own projects.
