Step 1 Start camera  |  Step 2 Click Start capturing for each class → click again to stop  |  Step 3 Collect 20–30 samples per class — keep counts roughly equal  |  Step 4 Click Train model  |  Step 5 Save model then switch to Monitor tab. Tip: for Patch reversed, show the patch clearly upside-down.
📷 Training camera
Off
● CAPTURING
Start camera to begin
⚙ Training controls
Training progress
StatusIdle
Step—
Epoch—
Loss / Acc—
Start camera, then capture samples per class.
🏷 Detection classes
📋 Train log
SERVER MODE — AI runs on the Python server (PyTorch, much more accurate). Place patch in guide box → Scan patch or enable Auto 3s. Server URL: —  |  ← Back to Train
Scanned
0
frames
Passed ✓
0
correct
Errors ✗
0
failed
Inference
—
server ms
📷 Camera input
Off
Start camera to monitor
Resolution: Facing:
🔌 Server
—
🤖 AI prediction (server)
Waiting
📋
Place patch in frame
Start camera and scan
Class breakdown
📋 Monitor log
⚠
Patch Error
Problem detected
Error
Class: — Conf: — Time: —