[SPIRIT-OC2] AIVATAR: Test Datasets

Fluendo proposes harnessing its expertise in advanced video codecs, AI-driven multimedia enhancements, and real-time video processing to contribute towards immersive learning within the SPIRIT platform, building on its innovative products (Fluendo Codec Pack, Fluendo AI Plugins, and Raven AI). Fluendo offers a proven track record in scalable, resource-efficient solutions for multimedia content delivery. The Experiment objectives include: Integrate LCEVC as a codec enhancement layer: Implement LCEVC as an enhancement layer, designed to improve the performance of any base codec. For this experiment, LCEVC will be applied on top of AVC (H.264), achieving up to a 20% BD-Rate reduction. This approach demonstrates the scalability and adaptability of LCEVC, tailored to real-world bandwidth constraints and enhancing the efficiency of video transmission. Develop AI-driven visual enhancements at the edge: Enhance fine details in avatars, including facial expressions, textures, and environmental elements, using AI-powered superresolution to improve clarity at higher display resolutions up to 4K. Demonstrate real-time integration: Utilize pre-trained, lightweight AI models optimized for resource-constrained endpoint devices, ensuring minimal latency (<200 ms) and seamless operation within SPIRIT’s "Real-Time Animation and Streaming of Realistic Avatars" use case. Benchmark against existing solutions: Conduct empirical performance evaluations to validate LCEVC’s enhancements over standard implementations and assess the impact of AI-driven improvements on realism and user experience. This experiment directly supports SPIRIT OC2’s goals by validating third-party applications and introducing scalable, low-latency enhancements tailored for immersive telepresence.