Demos — commands, real output, what each one proves
Eight reproducible demos. Every one lists prerequisites, the exact command, the output we got, what it proves, what it does not, and how it fails.
Event 2026-09-27Published 2026-09-27SE Inspect build 1.9.37
DEMOS 06
Run the alpha without hardware
laptop · 5 min
OPEN →DEMOS 00
Device discovery
Pi 5 + card · 2 min
OPEN →DEMOS 01
Virtual mapping and the M3 head
laptop · 3 min
OPEN →DEMOS 02
Physical mapping with negative controls
Pi 5 + card · 5 min
OPEN →DEMOS 03
On-chip learning and “Why recognised?”
Pi 5 + card + camera · 15 min
OPEN →DEMOS 04
Train the M3 detector (smoke run)
training env · 10 min
OPEN →DEMOS 05
Segmentation benches: vignetting and glare
laptop · 2 min
OPEN →DEMOS 07
Line state machine and card watchdog
laptop · 1 min
OPEN →DEMOS 08
Reference pipeline without a network
laptop · 3 min
OPEN →Each demo follows the same structure: Needs · Environment · Run · Output we got · Proves · Does not prove · Known failure modes · Time. Outputs were captured on 2026-09-27 from build 1.9.37; where a figure comes from a synthetic scene, the page says BENCH.
| # | Demo | Needs | Time |
|---|---|---|---|
| 06 | Run the alpha without hardware | laptop, Python 3.11 | 5 min |
| 00 | Device discovery | Pi 5 + card | 2 min |
| 01 | Virtual mapping and the M3 head logic | laptop | 3 min |
| 02 | Physical mapping — with negative controls | Pi 5 + card | 5 min |
| 03 | On-chip learning and "Why recognised?" | Pi 5 + card + camera | 15 min |
| 04 | Train the M3 detector (smoke run) | training environment | 10 min |
| 05 | Segmentation benches: vignetting and glare | laptop | 2 min |
| 07 | Line state machine and card watchdog benches | laptop | 1 min |
| 08 | Reference pipeline without a network | laptop | 3 min |
Images on the demo pages are produced by the scripts themselves (VORSA_BENCH_BILDER=<dir> for the benches, --out for the reference pipeline) — you can regenerate every figure.