
In modern cricket, run-out and stumping decisions rely heavily on the Third Umpire manually reviewing video footage frame-by-frame. This process is time-consuming, disrupts match flow, and can still lead to errors in close calls under pressure. To address this, we built AI 3rd Umpire—an intelligent cricket review assistant that streamlines the decision-making process using multimodal AI. Powered by Google Gemini, the system analyzes video footage to quickly identify key moments and assist in making accurate decisions. The application follows a two-step pipeline. First, using Gemini’s video understanding capabilities, it processes the uploaded clip to automatically detect the critical event—such as the moment the bails are dislodged—and jumps directly to the relevant frame. Next, OpenCV extracts a sequence of surrounding frames, which are then analyzed by Gemini’s vision model. The AI evaluates spatial relationships between the bat, crease, and stumps to generate a decision along with supporting reasoning. This pipeline is presented through a clean, broadcast-inspired Streamlit interface designed for clarity and usability. The system enables faster review workflows while reducing manual effort, offering a more efficient and assistive approach to third umpire decisions.
19 May 2026