← sources

I’ve been building a padel analytics pipeline from one fixed camera

source: reddit (reddit:computervision)published 2026-10-13 18:10ingested 2026-10-14 02:37

retained text

this video is auto-generated highlights from a 19 min amateur padel match, filmed with one GoPro low behind the baseline. The pipeline has player tracking, ball tracking ,picks, cuts and annotates the rallies on its own

YOLO11 tracks the players, a fine-tuned YOLO11 model i built finds the ball, and a lens-corrected court calibration turns positions into metres. Hits and bounces come from turns in the ball’s path and the card after each rally shows player heatmaps and where the shots landed.

Still rough: the ball is only found on about a third of the frames, so some hits get missed and speeds are estimates, not radar

if i'm to build a mobile product of this, how can i make the latency better - i mean , now when i upload the video player tracking, ball tracking, highlight generations and everything in the pipeline takes about 40min to 1hr based on the length of the video but the actual user wont wait that long

so what can i do to make the latency better (please dont say buy gpu's - im just a engineering student and im into CV,ML,DL stuff like that and im building this as a project)

i'm open to any advices on how to make it better!!!

document fetched 2026-10-14 00:23 · canonical https://reddit.com/r/computervision/comments/1x544li/ive_been_building_a_padel_analytics_pipeline_from