SwimVision LM is a cutting-edge, AI-powered web application that helps swimmers and coaches analyze swimming technique through automated, intelligent video analysis. Leveraging state-of-the-art Vision-Language Models (VLMs) like GPT-4o, SwimVision can break down swimming movements, identify key technique issues, and deliver tailored improvement suggestions—all by simply uploading a video. The platform is designed to democratize expert coaching, enabling swimmers of all levels to benefit from professional-level feedback, anytime and anywhere.
-
Video Upload & Frame Sampling
Upload swimming videos directly through the web interface. The system automatically samples relevant frames for analysis. -
AI-Powered Technique Analysis
Utilizes advanced VLMs to recognize swim strokes, body posture, and movement phases. Provides detailed, actionable feedback in natural language. -
Key Frame Preview
Visualize and browse the automatically selected key video frames used for analysis. -
Actionable Improvement Suggestions
Get customized recommendations for refining swimming technique, based on industry standards and best practices. -
Session History Management
Efficiently manages analysis results and previously uploaded videos for progress tracking. -
Simple Web Demo via Gradio
User-friendly web interface powered by Gradio—no installation or technical skills required. -
Extensible & Modular Codebase
Well-structured, modular Python code for easy development, scaling, and integration with additional AI or computer vision components.
- Upload a swimming video from your device.
- The system will process the video, extract key frames, and analyze them with an AI model.
- Receive detailed technique analysis and practical training suggestions.
- Explore your session history and follow your progress over time.
conda create -n swim-vision-lm python=3.11.7 -y
conda activate swim-vision-lm
pip install -r requirements.txt
gradio main.py