VideoCognition

VideoCognition Towards Explainable Alarms
Realization
  • Houda Chabbi
  • Jean Hennebert
  • Vincent Audergon
  • Oussama Zayene
  • Killian Ruffieux
  • Lucas Genoud
  • Valentin Biolley
  • Luca Michel
Keywords
  • Vision Language Model (VLM)
  • Large Language Model (LLM)
  • Artificial intelligence
  • Agentic AI
  • Microservice
Valorization
This project advances video analysis with AI-driven real-time event detection and intelligent forensic search capabilities.
Partners

Our Partners:

  • HEIA-FR
    • SeSi
    • iSIS
    • iPrint
  • EPFL
  • CSEM
Funding
Morphean SA
Schedule
03.2023 – 01.2026

Morphean offers an intelligent, scalable, and secure cloud platform that enables clients to monitor video surveillance feeds in real time. Their system takes a proactive approach to security by automatically detecting events of interest as they occur.

This project aims to extend Morphean’s existing toolset by introducing two key features. 

The first is a real-time and reliable alarm system designed to help security agents gain instant insights. It leverages Vision-Language Models (VLMs) to detect relevant events in live video feeds that may require human intervention.

The second feature focuses on forensic video analysis, providing an agentic AI assistant capable of helping users search through existing video archives. This assistant understands the user’s intent and dynamically builds the most effective search pipeline using the appropriate video analysis tools to deliver accurate and relevant results for the user query.