Sports analytics · Computer vision
Player analysis at broadcast speed.
Technical partner of a Saudi football analysis product · In production
REC · MATCH FEED · CAM 02
EVENT · progressive pass · #7 → #10
22/sec
Events tracked
96%
Detection accuracy
<2s
End-to-end latency
01 · The challenge
Broadcast footage, structured data.
Analysts needed every player, every event and every pass from standard broadcast footage — fast enough to use at half-time, not the next morning.
[Full challenge narrative pending partner approval.]
02 · Approach
Four models, one pipeline.
Detect & track
Identity-stable tracking for all 22 players through occlusion and camera cuts.
Read the shirt
Jersey-number OCR ties each track to a named player.
Tag the events
Action models label passes, shots and duels as they happen.
Ship the insight
Heatmaps and post-match analytics land in the analyst's dashboard.
“[Partner executive quote — pending approval.]”
[Name] · [Title], [Partner]
Start a project
Have footage that needs understanding?
A 30-minute discovery call. You bring the workflow, we bring a plan and a timeline.
- Reply within 1 business day
- 30-min call · No pitch deck