Problems in the Current Market

Many companies claim that they can track basketball players' movements in real time using only vision cameras. However, in reality, there is no service that provides fully automated vision-only movement tracking without human intervention. In practice, player movement is tracked either by using wearable sensors or by involving human experts in the vision analysis process.

More specifically, the reason why purely vision-based MOT (multiple object tracking) technology is difficult to apply to basketball games is that, within the camera's field of view, players frequently overlap and occlude each other. As a result, during the process of associating players detected in the current frame with players being tracked from previous frames, large errors occur.

Current Market Performance
Based on BoT-SORT and ByteTrack, AssA (Association Accuracy) is around 33% on the TeamTrack Dataset / Kaggle 2024.
Occlusion Issues
Players frequently overlap, causing tracking failures
Manual Intervention
Current solutions require human experts for accuracy
Low Accuracy
Traditional methods achieve only 33% association accuracy
CourtLenz solves this problem with its patented fused-association technology.
The fused-association technology of CourtLenz identifies players who are occluded by other players among
those detected in the current frame, and adjusts their bounding boxes accordingly. Furthermore,
this technology accurately detects court lines in game footage, and based on these court lines,
places players’ bounding boxes onto a tactical board in BEV (bird’s eye view).
The fused-association technology improves AssA by fusing not only the association based on bounding boxes in game footage, but also the association based on positions on the tactical board in BEV.
0%
As described above, CourtLenz’s fused-association technology works well even with a single camera, but it also provides synchronized association between multiple cameras capturing the court from different angles. In this case, AssA is improved.
0%

Technique - Courtline Recognition & Warping

Our advanced computer vision algorithms precisely identify and track basketball court boundaries in real-time, enabling accurate spatial mapping and perspective transformation for enhanced game analysis.

Real-time Court Detection
Precision Tracking
Accurate court line detection with sub-pixel precision
Real-time Processing
Live analysis with minimal latency
Multi-angle Support
Works with various camera positions and angles
3D Mapping
Transforms 2D footage into tactical 3D views

Technique – Fused Association

Our revolutionary Fused Association technology maintains accurate player identification even through severe occlusions and overlaps, ensuring continuous and reliable data collection throughout the game.

Advanced Player Tracking

Usage

CourtLenz seamlessly integrates into the existing workflows of coaches and analysts, providing them with powerful real-time and post-game analytics. These insights are then made easily accessible through our intuitive interface, empowering players to track their own progress and parents to support their child's journey.

Intuitive Dashboard
Real-time Analytics
Export Capabilities
We are advised by Professor Younghyun Kim of Electrical and Computer Engineering at Purdue
University on our innovative technology, enhancing the technical completeness and reliability of our products.