Getting Started

Go from sensor data to a deployed edge AI model in 5 minutes.

1. Create an account

Sign up at luviner.com. The Explorer plan is free forever — no credit card required. You'll be able to train models and see accuracy, latency, and memory metrics immediately.

2. Prepare your data

Luviner accepts CSV files with sensor readings. Each row is one sample, each column is a feature, and the last column is the label (class).

# vibration_sensor.csv accel_x,accel_y,accel_z,gyro_x,gyro_y,gyro_z,temperature,rms_vibration,label 0.12,0.98,-0.05,0.01,-0.02,0.00,45.2,0.34,normal 0.15,0.97,-0.06,0.02,-0.01,0.01,45.5,0.38,normal 2.34,1.87,0.95,0.45,-0.67,0.23,67.8,4.56,bearing_fault 3.12,2.45,1.23,0.78,-0.89,0.34,72.1,6.78,misalignment
See CSV Format for detailed requirements, tips on data quality, and a downloadable example dataset.

3. Create a project and upload

  1. Go to Projects in your dashboard
  2. Click New Project — give it a name (e.g. "Motor vibration classifier")
  3. Upload your CSV file
  4. Luviner auto-detects features and classes from your data

4. Train your model

Click Train. Luviner automatically:

  • Selects the optimal network architecture for your data
  • Trains an optimized neural network tailored to your data and target hardware
  • Reports accuracy, confusion matrix, inference latency, and memory footprint

Training typically takes 10-60 seconds depending on dataset size.

Explorer plan: you can train and see all metrics, but binary export requires a Builder plan or higher.

5. Export and deploy

Once you're satisfied with accuracy (Builder plan required):

  1. Select your target architecture (ARM Cortex-M4, ESP32, RISC-V...)
  2. Add the UIDs of your target chips (manually or via CSV upload)
  3. Click Compile — Luviner generates a UID-locked static library
  4. Download .a (library) + .h (header)
  5. Include them in your firmware project and call the inference function
See Firmware Integration for step-by-step instructions on including the library in your build system.

What's next?