Getting Started with Levy Vision

This guide walks you through turning Levy Vision on for a single pilot subaccount and getting your first events flowing. Most operators start with helmet selfies and parking-pose validation — those work on any vehicle, with no new hardware. Sidewalk detection comes later, once vendor hardware is installed.

Plan for about 15 minutes of dashboard work, plus a quick test ride.

Before you start

Confirm you have:

• Admin role in the subaccount you want to pilot. You need write access to the safety settings page.
• A few active vehicles in the pilot subaccount so you can take a real test ride.
• An iOS or Android phone with the Levy rider app installed and signed in.
• (Optional but recommended) A second subaccount as your control — leave Levy Vision off there so you can compare.

Step 1 — Open Safety settings

1. Open the dashboard at fleets.levyelectric.com.
2. Pick the pilot subaccount from the subaccount switcher in the top-left.
3. In the sidebar, open Safety > Safety settings.

If you don't see the Safety group in the sidebar, your role may not have access yet. Ask a Super Admin to grant you safety permissions on this subaccount.

Step 2 — Enable helmet verification

In the Helmet verification section:

1. Toggle Helmet check enabled to on.
2. Leave Strict mode off for your first week. With strict mode off, a low-confidence inference fail logs as ambiguous and still lets the ride start — which limits false-positive friction while you tune.
3. Leave Max attempts at the default of 3. Riders get three tries before the unlock is blocked (or, with strict off, logged).
4. Click Save.

That's it. The next time a rider opens the app to unlock a vehicle in this subaccount, they'll see the new helmet selfie screen.

For details on what riders see and how to read results, jump to Helmet verification at unlock.

Step 3 — Enable parking-pose validation

In the Parking-pose validation section:

1. Toggle Parking validation enabled to on.
2. Classifier — leave at Gemini unless you have a Captur.ai agreement. Gemini is included; Captur.ai is a premium add-on. See Vendor setup for the Captur option.
3. Reparking fee — set the amount in cents that gets applied for a high-confidence fail (default 500 = $5.00). Set to 0 to log fails without charging.
4. Click Save.

Riders will keep taking the same end-of-ride photo they always have. The classifier now grades it and writes a row to Parking reviews for any fail.

Step 4 — Take a test ride

1. Sign into the Levy rider app on a phone where you've been added to the pilot subaccount.
2. Find a vehicle in the app, tap Unlock.
3. You should see the new Helmet selfie screen. Take a selfie with a helmet on; the ride should start.
4. Ride a short distance, then end the ride and take the parking photo as normal.

Now back in the dashboard, go to Safety > Events. You should see at least two new rows — one helmet_check (status pass) and one parking_pose (status pass or fail depending on how the photo came out).

Step 5 — Flip strict mode on (when ready)

Run your fleet with strict mode off for a week and watch the Helmet compliance page. Look at:

• Pass rate — should be 90%+
• Ambiguous rate — should be under 5%
• Confirmed false positives (from operator-flagged events)

When you're comfortable with the accuracy, go back to Safety settings, toggle Strict mode on, and save. From that moment, a high-confidence helmet fail blocks the unlock after the max-attempts retry ladder.

Step 6 — Plan for sidewalk detection

Sidewalk detection requires on-vehicle hardware. The shortest path:

1. Read Sidewalk detection and Vendor setup.
2. Reach out to your Levy account manager to start the Drover AI or Luna Systems conversation.
3. Once devices are installed and provisioned in Safety > Vendor devices, turn on Sidewalk detection in safety settings.
4. Start with Throttle cut off so you only get events and rider warnings while you tune thresholds. Read Throttle-cut policy before enabling enforcement.

Step 7 — Set up MDS export (cities only)

If you bid on city permits, give your compliance contact a feed:

1. Open Safety > MDS export.
2. Pick a date range and a jurisdiction polygon.
3. Download the CSV or copy the MDS-aligned JSON URL.

See MDS export for cities for full details.

What to do next

• Read Helmet verification to understand the rider flow in detail.
• Skim Reviewing parking appeals so your team knows what to do when a rider disputes a fee.
• Bookmark Troubleshooting for the first time something looks off.