Team Security Guardz

Inspiration

We wanted to turn human intent into art without relying on rigid, predefined gestures. The core idea: the same motion can feel different each time. Instead of classifying gestures, we built a system that interprets intent.

How We Built It

We captured 6-axis IMU data using an ESP8266 + MPU6050, then filtered and oversampled the signal. From this data, we extracted motion features:

• Path geometry
  
• Curvature
  
• Speed profile
  
• Energy
  
• Smoothness
  
• Symmetry
  
• Dimensionality
  

These features are sent to Gemini for interpretation and prompt generation, which then drives image creation.

The web interface streams in real time:

• Motion trace
  
• Extracted features
  
• AI interpretation
  
• Generated artwork
  

What We Learned

• Converting noisy sensor data into stable, meaningful features
• Structuring prompts so Gemini explains its reasoning
• Designing a pipeline where ambiguity is a feature, not a bug
• Balancing smoothing with responsiveness to preserve the feel of motion

Challenges

• Filtering and calibration without removing expressive motion
• Keeping AI interpretation consistent and explainable while allowing creative variance
• Ensuring graceful degradation when image generation is unavailable

Why It Matters

Air-to-Art Sculptor transforms real-world movement into instant visual feedback. A simple gesture is captured by sensors, interpreted by AI, and turned into meaningful visuals in real time.

This can power interactive museum exhibits where visitors ‘paint’ with motion, or physical therapy sessions where movement quality becomes motivating visual feedback. Every gesture is detected, the reasoning is explainable, and the result appears within seconds.

This project treats motion as a continuous expressive signal. It converts physical intent into visual output, showing how AI can augment creativity instead of restricting it.

We define a conceptual success metric:

$$ I = \frac{\text{expressiveness}}{\text{rigidity}} $$

Goal: maximize interpretive confidence while preserving creative diversity.

Built With

• arduino
• arduinojson
• c++
• css
• esp8266httpclient
• esp8266httpclient-ai/apis:-google-gemini-api
• esp8266wifi
• flask
• gemini-3-preview
• html
• html/css/javascript-(web-ui)-frameworks/libraries:-flask
• javascript
• json
• mpu6050
• platformio
• python
• python-(backend)
• stability-ai-(optional-image-generation)-hardware/platforms:-esp8266-(nodemcu/wemos-d1-mini)
• stabilityai