About Revisor

The Revisor system is a neural network-based software package for counting the number of actual voters and monitoring compliance with some electoral procedures. This AI-enabled monitoring is fast, reliable, inexpensive and doesn’t know the limits of the scale. For a fraction of the cost, you can deploy an election observation mission with thousands of poll watchers covering 100% of polling stations in the target constituency. 

And those poll watchers don’t even blink!

The system utilizes the ability of neural networks to learn to recognize physical objects, track their relative movements, detect voting events and distinguish them from other activities at a polling station. The system can independently count the number of voters with high precision (up to 98% accuracy, depending on the relative position of cameras and ballot boxes in a polling station). 

The Revisor system is a trainable neural network, so the customer can teach it to work with different types of voting procedures, elections, and electoral systems in any country. Operation based on video recordings allows results to be obtained immediately after an election and/or months and years later.

Development stages 

Stage 1. December 2019 – July 2020 — Prototype 

Stage 2. August 2020 – September 2021 — System finalization 

Stage 3. October 2021 – Current time Multiple voting detection module

What can Revisor do?

• To detect ballot boxes on video records
• Detect a ballot box’s outline and its top
• Identify the type of the ballot box
• Identify parameters of box location
• Detect some types of violations

To count turnout

• Count the number of voters who cast their ballots
• Identify polling stations with falsified turnout comparing the real turnout to the officially published results
• Draft a formal complaint
• Speed up manual recount

How does Revisor work?

Results

17 million stuffed ballots

At the request of Meduza’s investigative team, Revisor processed more than 1 million hours of video records from Duma elections held in 2021. 

The revelation of 30% of falsified votes in the provided sample led the journalists to the conclusion regarding the whole country. 

Success stories

Revisor detects the discrepancy between the official and actual turnout at the polling stations and reports it to the users. It’s the users’ responsibility to resolve what particular crime led to the observed anomaly.

Occasionally, the crimes themselves are caught on video. So, when users watch the records indicated by Revisor as suspicious it gives us the ideas on how the culprits committed the crimes and formed the discrepancy.

Contact us

For advice on the feasibility of using the Revisor system at your particular election and for recommendations on setting up cameras, obtaining and storing recordings, please contact:

info@pollwatcher.ai

+ 370 61 33 88 61