Building an AI-Powered Drone Footage Pipeline
Independent Consulting
Situation
Drone operators spend hours manually reviewing flight footage to find the good shots — interesting subjects, proper exposure, stable frames. With the rise of commercial drone operations, this manual review process does not scale. I saw an opportunity to combine my growing AI skills with the drone industry.
Task
Build a production CLI tool that could automatically analyze drone footage using computer vision and AI, identifying highlights, detecting objects of interest, and generating organized output — turning hours of review into minutes.
Action
Built Skyforge as a Python CLI using YOLOv8 for object detection, CLIP for scene classification, and Claude/GPT-4o for intelligent scene description and highlight ranking. The tool processes video files, extracts frames at configurable intervals, runs multi-model analysis, and outputs organized results with confidence scores. Used Rich for beautiful CLI output and Click for argument parsing. Designed the architecture to be model-agnostic so new vision models can be swapped in.
Result
Skyforge processes drone footage automatically, identifying key moments and generating descriptions that would take a human operator hours to produce manually. The project demonstrates the intersection of AI and drone operations — exactly the kind of tooling that commercial drone companies need as they scale.