The IISc interdisciplinary AIM@IISc team with the IEEE TCSC SCALE Challenge 2026 award
A team from the Indian Institute of Science (IISc), Bengaluru, has won the IEEE Technical Community on Scalable Computing (TCSC) SCALE Challenge 2026, held alongside the IEEE/ACM International Symposium on Cluster, Cloud, and Internet Computing (CCGrid) in Sydney in May 2026. The award recognises the team’s entry, “Scaling Real-Time Traffic Analytics on Edge–Cloud Fabrics for City-Scale Camera Networks,” which addresses the challenge of processing hundreds to thousands of CCTV streams for traffic analytics under stringent latency, bandwidth, and compute constraints.
Prof Yogesh Simmhan (IISc) receiving the IEEE TCSC SCALE Challenge Award at IEEE/ACM CCGrid Symposium, Sydney, 2026
The IEEE TCSC SCALE Challenge is a long-running international competition, now in its 19th edition. It highlights real-world systems capable of scaling across multiple dimensions: scale-up, scale-out, and elastic edge–cloud computing, validated through technical evaluation and live demonstrations. Finalists present working deployments at the CCGrid symposium, where a panel selects a winner based on scalability impact and system innovation. The 2026 finals featured entries from teams in the United States, Italy, and Australia, besides India, covering domains such as big data architectures, edge-cloud systems for assisted living, and HPC-based parallel graph optimisation, underscoring the breadth of scalable computing research showcased at the event.
Live multi-stream detection and tracking on heterogeneous edge accelerators using the AIITS pipeline
The IISc team’s winning system presents an AI-driven Intelligent Transportation System (AIITS) that transforms multi-camera video feeds into a dynamic, city-scale traffic dashboard. Video streams from distributed cameras are first processed at edge accelerators using Deep Neural Networks (DNNs) for vehicle detection and tracking, producing lightweight summaries of traffic flow. These summaries are then aggregated in the cloud and analysed using Spatio-Temporal Graph Neural Networks (ST-GNNs) for real-time traffic nowcasting and short-term forecasting. This edge-first design reduces bandwidth requirements while enabling low-latency, scalable analytics for urban mobility applications.
SAM3-assisted labelling and federated fine-tuning orchestrated across hundreds of containers
Demonstrated on a large-scale testbed emulating a 400-camera deployment in a Bengaluru neighbourhood, the system integrates Raspberry Pi-based video ingestion with NVIDIA Jetson edge accelerators for high-throughput inferencing. A central contribution is a capacity-aware and energy-aware scheduler that dynamically distributes video streams across heterogeneous devices to sustain real-time performance at scale. The platform further incorporates foundation model-assisted labelling and continuous federated learning, enabling the traffic detection models to adapt to evolving road conditions and vehicle types without requiring centralised retraining or raw data transfer.
Cloud-hosted Spatio-Temporal GNN nowcasting/forecasting with interactive dashboard visualisation
The work was carried out by the interdisciplinary, AI for Integrated Mobility (AIM@IISc) consortium, involving faculty, students and staff from the Department of Computational and Data Sciences (CDS), the Centre for infrastructure, Sustainable Transportation & Urban Planning (CiSTUP), the Robert Bosch Centre for Cyber Physical Systems (RBCCPS), and the Centre for Data for Public Good (CDPG) at IISc. Supported by funding from CiSTUP and ARTPARK and in close cooperation with the Bangalore Traffic Police, the project demonstrates how scalable AI systems deployed across edge–cloud fabrics can address pressing urban mobility challenges, including traffic congestion, road safety, and sustainability in rapidly growing mega-cities like Bengaluru.
REFERENCE:
Scaling Real-Time Traffic Analytics on Edge-Cloud Fabrics for City-Scale Camera Networks, Akash Sharma, Pranjal Naman, Roopkatha Banerjee, Priyanshu Pansari, Sankalp Gawali, Mayank Arya, Sharath Chandra, Arun Josephraj, Rakshit Ramesh, Punit Rathore Anirban Chakraborty, Raghu Krishnapuram, Vijay Kovvali and Yogesh Simmhan, in 2026 IEEE 26th International Symposium on Cluster, Cloud and Internet Computing Workshops (CCGridW), TCSC SCALE Challenge Winner, 2026 (DOI 10.1109/CCGridW69005.2026.00027)
