REVISOR — Your Unblinking Poll Watcher

REVISOR is a neural network-based software package designed to count actual voters and monitor compliance with electoral procedures. It enables fast, reliable, and inexpensive AI-enabled election observation, capable of scaling to thousands of poll watchers covering 100% of polling stations in a target constituency.

The system uses neural networks to recognize physical objects, track movements, detect voting events, and distinguish them from other activities at polling stations. It can independently count voters with high precision (up to 98% accuracy, depending on camera and ballot-box arrangements). REVISOR is trainable, allowing adaptation to different voting procedures, elections, and electoral systems across countries. It relies on video recordings, enabling results to be obtained immediately after an election or long after it concludes.

---

How REVISOR Works

• Detects ballot boxes on video records.
• Detects a ballot box outline and its top.
• Identifies the type of ballot box.
• Identifies parameters of box location.
• Detects certain types of violations.
• Counts turnout by counting voters who cast ballots.
• Identifies polling stations with discrepancies between real turnout and officially published results.
• Drafts formal complaints.
• Speeds up manual recount processes.

Development & Deployment Timeline

• Stage 1 (Dec 2019 – Jul 2020): Prototype
• Stage 2 (Aug 2020 – Sep 2021): System finalization
• Stage 3 (Oct 2021 – Present): Ongoing improvements

Use Case & Impact

REVISOR has been used to analyze large video datasets (e.g., over 1 million hours of footage from elections) to reveal discrepancies between official turnout and actual turnout, highlighting potential falsifications. It provides users with indicators of suspicious activity and, when reviewed, can inform investigations. The system emphasizes that it is up to users to determine the specific criminal context behind observed anomalies.

If you’re considering deploying REVISOR, you can request feasibility advice, guidance on camera setup, storage, and overall deployment considerations via the contact information provided by the developers.

Core Features

• Neural network-based election observation at scale (thousands of poll watchers, 100% station coverage)
• High-precision turnout counting (up to ~98% accuracy, dependent on setup)
• Automatic detection of ballot boxes, their outlines, and types
• Monitoring of compliance with electoral procedures via automated event detection
• Violation detection and discrepancy analysis between real and official turnout
• Formal complaint drafting assistance
• Accelerated/manual recount support
• Trainable system adaptable to different countries, procedures, and systems
• Video-based operation with immediate post-election results or delayed analysis

---

How to Learn More / Contact

For feasibility assessments, camera setup guidance, data storage, and deployment recommendations, contact: [email protected] | +370 61 33 88 61

Proudly powered by WordPress