Organic vs Synthetic Produce AI
Edge AI system classifying produce in real-time using spectral sensor fusion and embedded ML models.
Chemical Integrity Automation
Distinguishing organic apples from chemically treated ones on an industrial conveyor belt manually is impossible. We created a spectrometer-driven ML sensor.
Spectral Analysis
Uses non-visible light spectrum scattering to determine chemical coating.
MCU Embedded
Model runs entirely on an ARM Cortex M4 microcontroller.
Offline Independence
No cloud connectivity required for classification.
Technical Strategy
Instead of large image arrays, we relied on high-frequency 1D spectral data to drastically improve processing time.
1D CNN Architecture
Designed a custom 1D Convolutional Neural Network specifically to analyze spectral wavelength peaks.
Edge Impulse Integration
Utilized the Edge Impulse studio to rapidly prototype and compile the C++ firmware package.
Automated Sorting Gate
Wired the GPIO outputs from the MCU directly to a pneumatic arm, physically sorting products in under 12 milliseconds.