EMERALD

Overview

EMERALD is a Horizon EU project partners including the British Broadcasting Corporation, Disguise Systems Limited and others in the area of Cinema post production and content delivery. I am the Co-PI of Emerald and our group leads significant work packages in this project. Our key contribution will be to enable granular monitoring and profiling of post production workflows, with an aim to improve the utilisation of AI and automation tools through efficient integration and utilisation of software-hardware accelerators.

Within TCD, this project will also work closely with the Sigmedia group who are involved heavily in optimising content streaming with strong links to Google (YouTube), NetFlix and others. Dr. Hareesh is leading the development of a granular energy measurement framework that will enable accurate profiling of energy consumption of existing post-production computing nodes (e.g., Compositing, Rotoscoping, Denoising). This paves way towards energy-aware development of applications and associated accelerators for sustainable media production in the future. Shashwat Khandelwal is leading the development of highly quantised deep-learning accelerators for high-resolution image manipulation workflows with PCIe-connected FPGA accelerators and their seamless integration with post-production tools such as Foundry’s NUKE.

News

• 2025/10/15: Paper “An Empirical Study of Reducing AV1 Decoder Complexity and Energy Consumption via Encoder Parameter Tuning” accepted by PCS 2025.
• 2025/10/14: Paper “Towards Energy Monitoring in Visual Processing Pipelines” accepted by SMPTE 2025.
• 2025/10/04: Paper “ReTiDe” accepted by CVMP 2025.

Objectives

In this project, RCSL was primarily responsible for the development of WP3 and WP5.

WP3

• An open infrastructure exposing energy measurement APIs to end-users and developers for visual postproduction flows
• Granular information through an “energy performance” tool enabling users to forensically probe their energy usage and performance bottlenecks
• Sophisticated set of tools to allow precise measurement of pipeline consumption locally, in the cloud, and during VP

WP5

• Re-use of hardware blocks under the control of ML algorithms to generate output streams at a quality suitable for postproduction, with low power consumption
• Generation of video features from compressed bitstreams without power

Phases

Publications

Vibhoothi, Vibhoothi, et al. “An Empirical Study of Reducing AV1 Decoder Complexity and Energy Consumption via Encoder Parameter Tuning.” arXiv preprint arXiv:2510.12380 (2025).

Hareesh Veekanchery, et al. “Towards Energy Monitoring in Visual Processing Pipelines.” Proceedings of the SMPTE Media Technology Summit. 2025.

Li, Changhong, et al. “ReTiDe: Real-Time Denoising for Energy-Efficient Motion Picture Processing with FPGAs.” Proceedings of the 22nd ACM SIGGRAPH European Conference on Visual Media Production. 2025.

Members

Acknowledgments

This work was supported by the Horizon CL4 2022, EU Project Emer- ald, 101119800.