contact us

Case Studies

Hockey Leagues Games Annotation

AI-Powered Data Annotation For Enterprise Machine Learning
TECH
SPECIALISTS
HOCKEY
LEAGUES
EVENT
MARKS
ACCURACY
RATE

The idea of the project

arrow-icon
Our Customer

One of the top providers in AI-powered data analysis with the use of data labeling solutions for enterprise-grade Machine Learning.

Leading in streaming, high-resolution, high-density video annotations, and in delivery of accurately-annotated, model-ready data to train and validate ML models.

arrow-icon
Project Goal

The client is developing an AI system to provide accurate hockey match statistics and analytical data to other companies.

Due to limited deep analytics extraction capabilities at this development stage, the client required more accurate AI training using manual labor and human expertise.

ENTERPRISE ML · AI-POWERED VIDEO ANALYTICS · SPORTS DATA LABELING

Solution

More than 100 professional tech-specialists annotated video recordings from 11 hockey leagues using bounding box and Skeleton technology.

Leagues Covered
USHL · AHL · NCAA · WHL · PWHPA · QMJHL · OHL · Allsvenskan · SHL (live) · Sweden J20 · SMSarja (J20)
FOW Analysis
Face-offs and whistle events captured and annotated with precision timing.
TOI Analysis
Time on ice for every player tracked across all game periods and shifts.
Eventing
40+ different event marks per match for complete game intelligence.

Key Features & Technologies

CAPABILITIES
  • Objects capturing and recognition
  • Objects attribution
  • Objects segmentation (FOW / TOI / Eventing)
APPROACHES
  • Bounding Box Technology
  • Skeleton Technology
TOOLS
  • OUTFORZ DATA custom annotation platform
  • Custom semi-automated annotation tools

Project Development Scheme

1
ANNOTATION
TASK FORMULA
Define annotation rules, taxonomy and quality standards for hockey events
2
DATA COLLECTION
& COHORT SCREENING
Source and select game footage from 11 leagues; filter for quality
3
ANNOTATION
GUIDELINE
DEVELOPMENT
Build detailed annotation guidelines covering all 40+ event types
4
ANNOTATION
PRODUCTION
100+ specialists annotate matches semi-manually with custom platform
5
CORPUS
ADJUDICATION
QA validation pass to ensure 98% accuracy across all output datasets
ANNOTATION AI TRAINING · ANNOTATION PRODUCTION · CORPUS ADJUDICATION