Offsides — Tactical Edge Detection

Where the market gets it wrong. Multimodal AI analyzes UEFA Champions League footage using YOLO + Qwen-VL 72B on AMD MI300X to detect mispriced prediction markets.

Scorecard: 3/5 correct edge calls | Model: Qwen/Qwen2.5-VL-72B-Instruct | Generated: 2026-05-07

Select Match

Architecture: YouTube highlights → Frame extraction → YOLO detection → Annotation (OpenCV) → Qwen-VL 72B reasoning (AMD MI300X via vLLM on ROCm)

How it works: For each upcoming match, the system analyzes the most recent 3 matches for both teams. YOLO detects player positions and ball location. OpenCV renders tactical overlays (defensive lines, compactness ellipses, team colors). Qwen-VL reasons over these annotated frames alongside stats and market odds to identify where the market may be mispriced.

Built for the AMD Developer Hackathon 2026 (Track 3: Vision & Multimodal AI)