Detecting a goal is easy. Knowing whether the ball actually crossed the line is a geometry problem. In this #TactiVision R&D module, I combine:

goal keypoint detection to anchor the goal geometry

goal segmentation to estimate the goal area

confidence filtering

geometric projection

polygon-based validation

temporal stabilization The main challenge is robustness: a single weak keypoint near the post can shift the projected goal line, so the system must decide which geometric information can be trusted before confirming a goal. This module is part of my broader football Computer Vision pipeline for detection, tracking, pitch mapping and tactical analysis