Erlan Labs is in early development. Nothing is available yet, and we aren't taking on customers, pilots or partners at this time.

Why the point was won

The cross-court drop caught the opponent leaning back after the clear. The lift that came back was short, and the smash went to the side they had left open.

  1. 1 Serve
  2. 2 Drop
  3. 3 Lift
  4. 4 Clear
  5. 5 Drop
  6. 6 Lift
  7. 7 Smash

Computer vision for racket sports.

We're building automatic line calls and shot tracking for badminton, and working toward models that understand every rally.

What we're building

Aiming for broadcast-grade line calls, without the broadcast budget.

Professional line-calling systems cost tens of thousands of dollars per tournament. We're developing a system that aims to do the job with a few cameras installed in a gym, so that one day every match on every court could get line calling, smash speed and other insights.

Everything below is in development and shown with prototype results and illustrations. None of it is available yet.

Prototype

Was it in?

Each landing is placed on the court in 3D and judged by the cork, the way the rules do, with a replay of the moment it touched down for close calls.

Prototype

How fast was that?

The shuttle is tracked from the moment it leaves the racket, so a smash can be measured at its real speed, not what is left of it at the floor.

In development

Placement drills.

Pick a shot and a target, and see where every one landed. The same landing detection behind our line calls maps every shot, so players could see their spread and track their accuracy session after session.

Planned

Track your progress.

Sessions saved over time, so players can watch their accuracy, smash speed and consistency change week by week.

Example: sample data, not a real player

Clears on target

83%

+41 pts since your first session

Top smash

287 km/h

+14 km/h this month

Sessions

10

Every Monday since 24 Aug

Clears on target, per session
0% 25% 50% 75% 100% 24 Aug14 Sept5 Oct26 Oct 83% 0% 25% 50% 75% 100% 24 Aug26 Oct 83%
Clears on target per session (sample data)
SessionDateOn targetPercent
12026-08-245 of 1242%
22026-08-316 of 1250%
32026-09-075 of 1242%
42026-09-147 of 1258%
52026-09-217 of 1258%
62026-09-286 of 1250%
72026-10-058 of 1267%
82026-10-128 of 1267%
92026-10-199 of 1275%
102026-10-2610 of 1283%

In development

Highlight reels.

Each session cut into a reel of the rallies. Only the play makes it in: no walking around, no picking up shuttles, no waiting between points. Then players could pick out their best shots and share them with their club.

Example: sample data, not a real player

Session45 min

Picking up shuttlesWater break

Your reel11 min 08 s

Best shots

  • Longest rally 31 s
  • Fastest smash 287 km/h
  • Closest call In by 4 mm

Designed for the venue.

  1. Install once

    The plan: cameras mounted in the gym, covering each court, so players have nothing to carry or set up.

  2. Just play

    The system is designed to recognize when a rally starts and ends, and to ignore shuttles from neighboring courts.

  3. See every call

    The goal: line calls and smash speeds for every rally, with replays of the close ones, ready when you step off court.

Research

From measuring the game to understanding it.

Line calls and speeds tell you what happened. We are building models that tell you why: who controlled the rally, what each shot was trying to do, and what you could have played instead.

  1. Prototype

    Tracking

    The shuttle in 3D from a few cameras, precise enough to call lines.

  2. In development

    Movement

    Players’ bodies in 3D, so a swing can be compared with an expert’s.

  3. Research

    Understanding

    Models that read a rally as a sequence of decisions, not a list of shots.

Questions we want it to answer

  • Where am I losing most of my points?
  • Which rallies turned the second game?
  • What does my opponent do when I lift to their backhand?
  • What should I have played instead of that cross-court drop?
  • Is my backhand clear getting deeper this season?
  • How do our doubles rotations break down under pressure?

Status

Where we are.

Erlan Labs is in early development. We're building and testing on our own recordings, and nothing is available to buy or use yet. We're not looking for customers, pilot venues, partners or investment at this time.

The visuals on this page are illustrations based on prototype results.