.png&w=828&q=75)
video captioning agent for AMD Hackathon Track 2. It downloads video clips from provided URLs, extracts 5 keyframes using OpenCV with quality scoring (sharpness, contrast, content richness), and sends them to Google's Gemma-3-12B-IT multimodal model via the Fireworks API. The agent generates captions in four distinct styles: formal, sarcastic, humorous_tech, and humorous_non_tech in a single LLM call. The pipeline handles varied video content (nature, urban, animals, people, sports, food, weather, technology) and outputs structured JSON. Built with Python, OpenCV, and httpx. Containerized via Docker for evaluation harness.
12 Jul 2026