Agentic Drone Surveillance | Real-Time Threat Assessment with LLMs and Computer Vision

Description

This video showcases a working demo of an Agentic Drone Surveillance system, built as a practical, end-to-end reference for real-time aerial monitoring using agent-based workflows, computer vision, and LLM/VLM-driven reasoning. The demo walks through how live video feeds are processed using real-time object detection and selectively escalated to deeper contextual analysis only when threat patterns emerge. It highlights how agentic workflows enable efficient, explainable threat assessment without unnecessary latency or cost. What’s demonstrated in the demo: - Agentic threat assessment workflows orchestrated using LangChain - Real-time object detection with YOLOv8 - Vision Language Model (VLM) for contextual scene understanding - High-speed LLM inference using Groq (LLaMA models) - Django backend for APIs, event handling, and orchestration - React frontend for live monitoring, alerts, and visual insights - Persistent storage of drone telemetry and incidents using SQLite This project is designed to serve as a practical reference for engineers exploring drone intelligence, agentic AI systems, real-time computer vision pipelines, and LLM/VLM-assisted decision-making. 🔗 Repository: https://github.com/Nishant2116/drone-surveillance-backend.git 🔗 LinkedIn: https://www.linkedin.com/in/nishant-deshmukh2116/