Girish Np@girish_np

QuadFrame Vision is an AI-powered video captioning and understanding platform built for the AMD Hackathon. The system automatically processes video URLs, intelligently selects the most informative frames, and generates accurate natural-language captions using Google's Gemma 4 Vision Language Model accelerated with AMD ROCm. Unlike traditional approaches that analyze every frame, QuadFrame Vision performs adaptive scene detection and key-frame selection to reduce computation while preserving important visual information. This significantly improves efficiency without sacrificing caption quality. The platform supports multiple caption styles including formal, sarcastic, humorous technical, and humorous non-technical, making it suitable for accessibility, media automation, social media content generation, education, and intelligent video indexing. The entire solution is packaged as a Docker container for one-command deployment and is designed to run efficiently on AMD Instinct GPUs with automatic CPU fallback when GPU resources are unavailable. The application accepts a JSON task file containing multiple video URLs and produces structured JSON outputs, making it easy to integrate into existing AI pipelines and enterprise workflows. Key features include intelligent frame extraction, automated video downloading, robust error handling with retry mechanisms, scalable batch processing, configurable inference settings, and production-ready architecture. The modular design separates video ingestion, frame analysis, AI inference, and output generation, allowing future extensions such as multilingual captioning, video summarization, visual question answering, and content moderation. QuadFrame Vision demonstrates how AMD ROCm and modern vision-language models can be combined to build scalable, efficient, and deployable multimodal AI applications that transform raw video content into meaningful structured information for real-world use cases.

To participate in the hackathon, simply click Sign up with AMD. If you are not already a member of the AMD AI Developer Program (ADP), you will need to create an account. ADP members can access AI tutorials, experts, and community support. In addition, new members can claim a free month of DeepLearning.ai pro, $100 in AMD GPU cloud credits, and $50 in Fireworks AI API credits. All hackathon participants will receive on hackathon launch additional compute and API credits for their projects. Ready to build what's next on AMD? Sign up with AMD below to secure your spot.

Join the next edition of our flagship AI hackathon — now featuring Google, AppliedAI, and Qdrant as official partners, with over $45,500 in prizes and exclusive partner challenges. After an incredible launch earlier this year, we’re back with the next edition of our AI hackathon series leading up to the /function1 AI Conference & Exhibition in Dubai. Tap into the power of AI to innovate, build, and take your AI skills to the next level in this high-impact hybrid experience. 📅November 14–19, 2025 • Nov 14–18: Collaborate and build online with AI enthusiasts from around the world. All projects must be submitted by end of day on November 18 • Nov 18 (Dubai): An exclusive on-site build day at Festival Arena Dubai for selected participants. • Nov 19 (Dubai): Live on-stage pitching & winners announcement at the /function1 Conference. 🌟 Get support from expert mentors throughout. 👥 Go solo or team up. 📍 Please note: On-site participation is by invitation only. Travel and accommodation will not be covered for selected participants. Further details will be shared with those accepted. 🧑🏻‍💻 Secure your spot now—sign up before the Kick-Off Stream!

🕒 6 days to explore the full potential of GPT-5 💡 Build next-gen AI apps 🤝 Compete solo or form a dream team 🚀 Connect with global innovators 🧑🏻‍💻 Sign up before the Kick-Off Stream to secure your spot!