TechniqueAI

AI-powered overhead press form analysis. Detects incorrect elbow and knee alignment in real-time to improve lifting technique and prevent injuries.

Advanced AI models analyze video input frame-by-frame, focusing on elbow flare, knee cave, and other common overhead press mistakes.

Read the full technical report (PDF)

Features

• Real-time detection of improper elbow and knee positioning during overhead presses
• Multiple model architectures: LSTM, Custom LSTM, Feedforward
• Video upload & keypoint extraction (MediaPipe)
• Frame-level predictions for knees and elbows
• REST API endpoint for integration with frontends/mobile apps
• Training pipelines with checkpointing and metrics tracking
• Evaluation metrics (accuracy, precision, recall, F1) per joint

Tech Stack

• Backend: Python, Flask
• Machine Learning: TensorFlow/Keras
• Computer Vision: MediaPipe, OpenCV
• Data Processing: NumPy, scikit-learn
• API: Flask + CORS
• Frontend: React, ChartJS, Tailwind