Match video processing
Detect players, ball events, court context, and movement evidence from real recorded padel clips.
Private pilot build by 42N
Padel match video, read as tactical intelligence.
WJETT helps players, coaches, academies, and clubs turn recorded matches into clear evidence: player movement, ball tracking, rallies, court zones, review clips, and performance insights.
WJETT product facts
- Stage
- Developing
- Runtime
- NVIDIA DeepStream
- Use case
- Padel analytics
Video to tactical report
- Rallies
- 42
- Review clips
- 8
- Runtime FPS
- 53.8
Coach insight
Pressure rises when the pair loses net position after the sixth shot. Review clips are ready for correction.
Product
A practical analysis layer for padel teams.
WJETT focuses on the decisions that happen after a match: what pattern repeated, where court position was lost, which clips explain the correction, and how the player or pair is evolving over time.
Coach-ready review
Convert raw video into reviewable rallies, clips, visual overlays, and reports that can be discussed with players.
Club and academy workflow
Build repeatable analysis for multiple players without forcing every coach to become a video engineer.
NVIDIA fit
Built around GPU-accelerated inference and model iteration.
The product uses NVIDIA technologies where they matter most: high-throughput video inference, TensorRT engine optimization, and production-scale validation for real court footage.
Runtime stack
- NVIDIA DeepStream as the promoted runtime path for product analysis.
- TensorRT FP16 engines for model acceleration and deployment validation.
- CUDA GPU infrastructure for inference benchmarking and model iteration.
- PyTorch checkpoints kept as reference models for parity and fallback evidence.
Representative local evidence
18 / 18 clips passed promotion gates
Current validation evidence tracks decoded detections, player identity stability, ball continuity, runtime artifacts, and full-pipeline speed before promoting a runtime path.
FinalF ball GT F1
0.7937
Raw tracker switches
0
DeepStream side-lane FPS
53.8
Pilot plan
Designed to turn GPU support into customer access.
Early GPU credits and cloud resources would let 42N validate production inference costs, process more representative matches, fine-tune detection quality, and offer lower-friction trials to the first clubs and coaches.
Production cost validation
Measure cost per match across realistic clip lengths, quality settings, and customer workloads.
Model improvement loop
Use larger GPUs for training, export, parity checks, and inference benchmarking without slowing down product development.
Accessible first pilots
Offer capped trials to coaches, academies, and clubs while the backend moves from private validation toward production.
Built by 42N
WJETT is a focused product, not a generic demo.
The current frontend is intentionally presented as a public product page while the backend is being hardened. The goal is simple: make the product understandable, credible, and easy to evaluate for NVIDIA Inception benefits.