NetSightGuided tour

Try the detection engine

See it find players and the ball.

Pick any clip on this device and press play. The same open YOLOX family NetSight builds on runs inside your browser — your video is never uploaded, which is exactly how we treat youth footage everywhere in the product.

Try it on your own clip

Create a free account to run live detection on your own video — it still happens entirely in your browser, nothing is uploaded.

Two real camera conditions

Both clips ran through NetSight's offline evaluation pipeline: YOLOX-S player detection fused with a dedicated volleyball ball tracker, foot-lock filtering, and ByteTrack keeping persistent player IDs through occlusions and missed frames. They show a school fixed-camera angle and a closer arena angle instead of presenting one favorable clip as universal accuracy.

School camera baseline

Wide, elevated fixed camera

West Washington Livestream · YouTube Creative Commons Attribution · 22-second excerpt · annotations added by NetSight.

What you're seeing (honestly)

This is the compact YOLOX-Nano model on generic COCO training — great at finding people, weaker on small fast volleyballs. The full product runs bigger models off-line (in NetSight Studio or a licensed worker), uses temporal and player-geometry gates to reject persistent shoe locks, turns detections into draft touch events, and always puts a coach between the AI and the published box score. Like what you see? Install NetSight Studio or request pilot access.