REVISOR is a sophisticated neural network-based software package designed to revolutionize election monitoring. Its core functionality lies in its ability to accurately count the number of actual voters and meticulously monitor compliance with various electoral procedures. This AI-enabled system offers a significant advantage over traditional methods by being fast, reliable, and remarkably inexpensive. For a fraction of the cost of traditional poll watchers, REVISOR can effectively deploy an election observation mission that covers 100% of polling stations within a target constituency.
The system leverages the advanced capabilities of neural networks to learn and recognize physical objects, track their relative movements, detect voting events, and distinguish them from other activities occurring at a polling station. This allows REVISOR to independently count voters with a high degree of precision, achieving up to 98% accuracy, depending on the camera placement and the configuration of ballot boxes. This level of accuracy ensures a robust and trustworthy count.
One of REVISOR's key strengths is its trainability. The neural network can be taught to work with different types of voting procedures, elections, and electoral systems across any country. This adaptability makes it a versatile tool for international election observation. Furthermore, its operation based on video recordings allows for results to be obtained immediately after an election, or even months and years later, providing a valuable historical record and a powerful tool for post-election analysis and accountability.
REVISOR's capabilities extend to detecting ballot boxes, identifying their outlines and types, and determining their location parameters. Crucially, it can also detect certain types of violations that might occur during the voting process. In terms of counting turnout, REVISOR can precisely count the number of voters who cast their ballots. This enables the identification of polling stations with potentially falsified turnout by comparing the real turnout data with officially published results. The system can assist in drafting formal complaints and can also speed up manual recount processes by providing accurate initial data.
The development of REVISOR has progressed through several stages, starting with a prototype in December 2019 – July 2020, followed by system finalization from August 2020 to September 2021. The current stage, from October 2021 onwards, focuses on enhancing the multiple voting detection module. The system's effectiveness is demonstrated through its ability to detect ballot box stuffing and other fraudulent activities, with video evidence often captured and analyzed by the system. REVISOR provides users with the discrepancy between official and actual turnout, empowering them to investigate the specific crimes that led to the anomaly. The system also offers a downloadable PDF for further details and can be integrated with various video sources, ensuring comprehensive coverage.
REVISOR is positioned as a powerful tool for election commissions, non-governmental organizations, and any entity concerned with ensuring the integrity and transparency of electoral processes. Its AI-driven approach offers a scalable, efficient, and objective method for safeguarding democratic elections.
