Choose a match video
Load an MP4 recorded from a stable camera. Full-court visibility and all four players improve analysis.
LOCAL AI VIDEO ANALYSIS · v0.2.4 · ENGLISH EDITION
Picklary Vision Rating
A Windows local video-analysis program with an English interface and English result reports. Analyze your own match video on your PC to receive Picklary DUPR estimates, a six-axis profile scored out of 7, and detailed shot, tactical, and movement insights.
EN
English UI · English result reportsThis English page provides the English edition only.
🔒 No server upload⚡ GPU first🧠 CPU fallback🗂 English UI & report
This is an independent Picklary estimate, not an official DUPR rating. Results depend on video quality, camera angle, and detection confidence.
RESULT PREVIEW · 7-POINT SCALE4.61Picklary DUPR
What shape is your game?Turn one match video into a clearly differentiated six-axis picture of strengths and priorities.
01 · HOW IT WORKS
From your video to an English performance report
Review the Windows local-analysis workflow through the quick-start guide and result example. When the executable package is available, this page will provide the download.
Confirm court and players
Review automatic detection and manually mark eight court points or player positions only when needed.
Run local analysis
The app prefers a DirectML GPU and automatically falls back to CPU. Your video is not sent to an external server.
Open the English report
The English edition uses English throughout the interface, analysis explanations, and result report.
ENGLISH EDITION
Windows local-analysis workflow
The current web deployment does not include the Windows executable ZIP. The quick-start guide and result example remain available, and this page will provide the download when the package is restored.
02 · RATING LOGIC
The score combines evidence, not just wins and losses
Ball, player, and court movement are combined with shot success, depth, tactical choices, movement coverage, and repeatability to build six dimensions scored out of 7.
Observed evidence
Collect ball and player trajectories, shot moments, court position, and rally segments from video.
Context classification
Classify serves, returns, drops, drives, net exchanges, and offense-to-defense transitions.
Confidence weighting
Give more weight to clearly detected evidence and reduce the impact of occluded or uncertain moments.
Profile and DUPR estimate
Combine the six-axis profile, video performance, and supplied context into a Picklary DUPR range and confidence score.
03 · RESULT EXAMPLE
See the shape of your game—not just one number
Axes extending outward indicate relative strengths; recessed axes reveal the next training priorities. Every skill dimension uses a 7-point scale.
Example Player A · 7-point scale
78% confidencePicklary DUPR 4.61
Return 6.6 / 7
An example where return depth, cross-court direction, and transition forward are consistently observed.
Offense 5.3 / 7
Combines drive speed, third-shot development, and attack-selection evidence.
Agility 3.8 / 7
Lateral coverage and depth response trail the other axes, making movement the next training priority.
Shots, court, and tactics
Includes shot types, speed, depth, court maps, positioning, strengths, and improvement suggestions with supporting evidence.
YOUR VIDEO · YOUR PROFILE
What radar shape will your match create?
Use the quick-start guide and result example to review the workflow. The Windows executable package will be available from this page when distribution resumes.
PICKLARY EDITORIAL
What Vision Rating can — and cannot — tell you
Vision Rating analyzes match video on the local PC and summarizes six axes: serve, return, offense, defense, agility, and consistency. The radar shape is meant to reveal a profile rather than reduce a player to one number.
It is not an official DUPR rating. Camera angle, occlusion, resolution, and the number of analyzable rallies affect confidence, so the report pairs estimates with confidence and observable evidence.
Source video stays on the Windows PC. The app prefers local GPU processing and falls back to CPU when needed. Use the result as a training-priority tool, not as a replacement for official match records.