AI counting system · Built for Hong Kong rope skipping
Camera + DGX Spark,speed counts on the floor.
Webcam + NVIDIA DGX SparkAI human trackingMulti-skill counting
We provide the webcam and NVIDIA DGX Spark setup on site for real-time inference. AI human tracking counts live — single rope, side-straddle, crosses, double unders and more, not only single-rope speed. Cuts manual tally load and error, with video kept for post-event review.
JumpFlow AI Counting
Automatic counting on site — near human-judge level
Webcam + NVIDIA DGX Spark kit, AI human tracking, and multi-skill support from single rope to double unders.
- 01
Webcam + NVIDIA DGX Spark kit
Hardware on site
We bring webcams and NVIDIA DGX Spark to your venue for real-time on-site inference — set up and run live. You don't need your own hardware or reliable venue Wi-Fi.
- 02
AI human tracking
Human tracking
Cameras track the athlete's body and foot paths in real time so you can see how each jump is counted — measured accuracy above 98%.
- 03
Single rope · side-straddle · crosses · double unders
Multi-skill
Not only single-rope speed — also side-straddle (開合跳), crosses, double unders and other common skills for championship and school events.
- 04
Multi-lane + video review
On-site + audit
One NVIDIA DGX Spark can process several lanes in parallel with faster-than-realtime results; every heat keeps video for dispute review.
JumpFlow AI Counting
Real hardware · Real accuracy
From the webcam + NVIDIA DGX Spark kit we provide, to AI human-tracking frames — clearly visible end to end, easy to review after the event.
01
Webcam + DGX Spark on site
Event-day inference runs on NVIDIA DGX Spark — webcam feeds, live count on screen.
02
AI human tracking
Annotated frames show body / foot paths and step-by-step counts — you can see how it tracks and tallies.
03
Predicted vs labeled
Test clips align Predicted with Ground Truth — errors are auditable.
JumpFlow AI Counting
Setup to score, in four steps
On event day we bring the kit, power on, and deliver live results — short and stable by design.
- 01
We set up webcam + DGX Spark
We provide webcams and NVIDIA DGX Spark, place them for your floor plan, map every lane to a camera input, and run real-time inference on site.
- 02
Ready the next heat
Operators mark the next heat ready on the console — then start time and lane mapping lock in.
- 03
AI human tracking counts live
Once athletes start, human tracking accumulates jumps — single rope, side-straddle, crosses, double unders and more; console or display updates in real time.
- 04
Sync scores + keep video
Results land when the heat ends; recordings stay available if you need a post-event review.
JumpFlow AI Counting
Built for rope-skipping counting on the floor
- Overall counting accuracy
- 98%+
- Average error (jumps)
- ±1
- Single · straddle · cross · double
- Multi-skill
- Webcam + DGX Spark on site
- Full kit
JumpFlow AI Counting
Next speed event — count with AI.
Includes webcam and NVIDIA DGX Spark setup for real-time on-site inference; quoted by event size. Short online or on-site demos available so your team can verify human tracking and multi-skill counting.