Back to the newsroom

Press Release

Korea Deep Learning selected to build 300 items of CCTV video data for intelligent monitoring

Korea Deep LearningPublished

Korea Deep Learning selected to build 300 items of CCTV video data for intelligent monitoring
In December 2024, Korea Deep Learning was selected as a contractor for NIA's Hyperscale AI Data Construction Project and will build 300 items of CCTV video data on safety incidents.

On December 3, 2024, Korea Deep Learning was selected as the contractor for the "CCTV Video Data for Intelligent Monitoring Services" project, part of the "2024 Hyperscale AI Data Construction Project" overseen by the Ministry of Science and ICT and run by the National Information Society Agency (NIA). Korea Deep Learning will build 300 items of CCTV video data capturing safety incidents and special events such as intrusions, falls, fights and crowding, laying the training groundwork for AI models that prevent and detect safety incidents.

Four types of safety incidents and two types of special events

The Hyperscale AI Data Construction Project builds large-scale AI training data to grow the AI ecosystem and widen the use of AI in everyday life. This project, one of its tasks, creates training data for AI models that prevent and detect safety incidents and can be applied to intelligent CCTV. Korea Deep Learning will secure CCTV footage from several institutions and build 300 items of video data covering four types of safety incidents, including intrusions, falls, fights and crowding, and two types of special events, including dense crowds and flooding.

Simplified tagging to lower the cost of building data

Turning large volumes of CCTV video into AI training data is costly. Korea Deep Learning applies a simplified tagging method to reduce the cost of building the data, and uses video analysis and anomaly detection technology to support the development of AI models that prevent and detect incidents. The finished data can be used in monitoring systems that detect dangerous situations in real time and respond quickly.

"We have been advancing vision AI technology that combines accuracy and security. We will work to build an intelligent monitoring system that detects dangerous situations in real time and responds quickly, so that AI can help build a social safety net."

— Kim Ji-hyun, CEO, Korea Deep Learning

NIA data construction projects are familiar ground

Drawing on its video understanding and object detection technology, Korea Deep Learning built high-quality video datasets that can be used in monitoring settings, and after finishing the build it also completed data quality verification.

NIA data construction projects are familiar ground for Korea Deep Learning. In the ink-wash painting data task of the 2022 "AI Training Data Construction Support Project," it was the first of 310 tasks to submit its results, exceeded its quantitative target by 235%, and received an "excellent" rating.

About Korea Deep Learning

Korea Deep Learning (KDL), founded in 2019, is an enterprise AI agent company. Built on KDL Frontier, its own vision-language model (VLM) specialized for documents, KDL makes AI agents that take on and complete the document work of businesses and public institutions. Its main products are DEEP OCR, DEEP Parser, DEEP Agent and DEEP Ops.

Source