Search results for ECCV
Telecommunications & Media
SKT, KAIST develop AI video compositing technology InsertAnywhere
SK Telecom and KAIST have jointly developed an AI video compositing technology called InsertAnywhere that can naturally insert products, props, characters or brand logos into existing footage. The system analyses 3D space and motion over time, including lighting, reflections and shadows, to blend images into a scene. SKT says it can cut tasks that took days to weeks down to hours and has automated more than 90 percent of the process. The technology was developed for SKT’s virtual PPL solution AdFlux, and the related paper was accepted to ECCV 2026.
Telecommunications & Media
SKT\'s InsertAnywhere accepted by ECCV, one of top three computer vision conferences
SK Telecom said its AI video compositing technology, InsertAnywhere, jointly developed with a KAIST research team, has been recognised by a major global computer vision conference. The company said a paper on the technology was accepted to ECCV 2026, to be held in Malmo, Sweden, from Sept. 8 to 12. InsertAnywhere adds objects not present at filming and maintains stable position, shape and lighting effects. SKT said it has commercialised the technology through its AdFlux solution.
Games & Commerce
Naver Labs Europe unveils \'DIVINE\' general-purpose encoder for autonomous robots, integrating visual AI models
Naver said it has introduced next-generation robot brain technology that integrates visual AI processing models for autonomous robots, enabling faster recognition of surrounding environments without heavy high-performance equipment. Naver Labs Europe unveiled DIVINE, a general-purpose encoder that converts data collected by cameras and LiDAR into a form AI models can process. It uses a multi-teacher distillation method to unify tasks such as 2D understanding, 3D reconstruction and human recognition, improving speed while reducing memory use.