The pipeline, as designed

Roadmap

The deployment design keeps the path from camera to insight short. Video is decoded and players detected and tracked; a pose estimator turns each player into per-frame keypoints; windows of those keypoints enter the foundation model, and one forward pass returns both outputs — a clean 3D skeleton and the movement embedding. No stage in the design requires special hardware on the pitch: the input is ordinary video.

videoordinary footagedetect + trackplayers per frameposekeypoints per playermodelone forward pass3D + 384-dskeleton + embedding
The designed serving path — one forward pass of the foundation model per movement window. Roadmap: not yet a production service.

Small enough to move

The backbone is a deliberately small model by foundation-model standards (see the architecture datasheet), and the deployment plan pushes further: post-training quantization to shrink memory and latency, and distillation into task-specific student models where a product needs only a slice of the model’s understanding. The goals are stated as goals — measured latency and throughput numbers will be published when a serving pipeline exists to measure them.

Design goals

Roadmap
GoalRationale
single-pass servingBoth outputs of the model from one forward pass — no chain of specialist models to operate.
quantizationPost-training quantization to cut memory and latency on commodity GPUs.
distillationTask-specific student models where a product needs a fraction of the backbone.
edge-class targetsAnalysis near the camera where bandwidth or privacy requires it.
batched pose front-endAmortized per-player cost when many bodies share one frame.

The query surface this pipeline is designed to feed is specified in the knowledgebase API.