Real Time Video Analytics

Real-time video analytics is a research line focused on the processing and analysis of video streams as they are being captured, enabling immediate interpretation and understanding of visual data without delay. These technologies employ advanced algorithms to process incoming video feeds instantaneously, allowing for swift detection and response to relevant events or patterns within the footage.

• Behaviour Analysis: AI-based analysis for scene-related and target-related behaviour recognition.
• Human Re-ID: AI-based analysis for human re-identification through visual appearance.
• Human Motion Understanding: AI-based analysis for spatio-temporal body pose estimation.
• Data fusion: Knowledge representation using graph-based storage including spatial, temporal and causal relationships.
• Risk Analysis: Inference and reasoning services for early detection of risk indicators.
• AI Model Monitoring: Methodologies for tracking, analyzing, and evaluating the performance and behavior of machine learningmodels in operational environments.
• AI Model Optimization: Advanced techniques for AI model optimization using Knowledge Distillation, Pruning, Quantization, Dynamic Batching or Model Architecture Redesign.