AI-Based Multiviewer for Network Disconnect Analysis

Objective: Develop an intelligent monitoring system that detects network disconnects across multiple streams and provides contextual analysis for troubleshooting and future model training.

Technologies Used: Python, OpenCV, TensorFlow, FastAPI, PGVector, and FFMPEG.

Challenges Addressed: Traditional network monitoring tools lack the ability to correlate visual stream data with network metrics, making root cause analysis difficult.

Solution: Created a multiviewer application that combines real-time video analysis with network telemetry. The system uses computer vision to detect visual anomalies during disconnects and stores this data alongside network metrics in a time-series database. An AI model provides contextual analysis of each disconnect event, categorizing them and suggesting potential causes and whats going with the content at the timestamp of the issue.

Outcome: Reduced mean time to resolution (MTTR) for streaming issues by a significant percentage through automated root cause analysis. The system's archive of disconnect events with contextual data provides valuable training material for improving future AI models. The multiviewer interface allows operators to monitor multiple streams simultaneously with color-coded alerts for different severity levels of disconnects.

Built using open-source computer vision and machine learning libraries.