The Hepta Insights Enhancements project, carried out between March and May 2025, successfully delivered a series of upgrades to the Hepta Insights platform, with a focus on video processing, annotation, and AI-powered defect detection.
As part of the video upload and processing improvements, the system now supports MP4 uploads of all sizes, including large files via chunked upload. Geospatial metadata is extracted from accompanying .srt files and correctly linked to the video content. All videos are securely stored in cloud infrastructure, fully integrated with Hepta Insights’ existing systems, and made accessible for playback. These features were deployed to both staging and production environments, with comprehensive documentation and a successful live demonstration marking their completion.
In parallel, a new video annotation system was developed. This allows users to pause video frames and draw bounding box annotations to mark defects. The annotated frames are saved as still images, enriched with metadata such as geolocation, and integrated seamlessly into existing reporting workflows, including exports to PDF and spreadsheets. This feature was tested, documented, and demonstrated to stakeholders, receiving full acceptance.
To enhance the platform’s AI capabilities, machine learning models were developed and trained to detect key infrastructure defects, including wire strand breakage, bird nests, vibration damper corrosion, and insulator disk shattering. All models achieved at least 70% precision and recall, were thoroughly documented, and deployed into the end-to-end processing pipeline within Hepta Insights. A live demonstration of model predictions was provided, and the solution was approved.
In May 2025, the project scope was extended to include the development of a new model for detecting non-compliant anti-bird spikes. This model was trained on a dedicated dataset, achieved performance metrics exceeding 95% for both precision and recall, and was successfully integrated into the same AI workflow. The final deployment and demonstration of this model concluded the project, with all objectives completed on schedule and to the agreed quality standards.
Importantly, the technical work was strengthened by valuable input from business development researchers. Their efforts in mapping and understanding market needs across the Central and Eastern European (CEE) region helped ensure that product enhancements were aligned with real-world demand and regional priorities.
Project Financing Information
-
Beneficiary name: Hepta Hungary Kft.
-
Project title: AI-Enhanced Asset Intelligence for Critical Infrastructure
-
Contracted grant amount: HUF 76,500,000
-
Aid intensity: 80%
-
Project completion date: 2025.05.31
-
Project ID No: 2023-1.1.3-STARTUP-2023-00028-3
-
Project summary: The project delivered advanced video processing, annotation, and AI-based defect detection features for the Hepta Insights platform. It was further supported by business development researchers who ensured alignment with infrastructure monitoring needs in the CEE region.
30.06.2025
