MoreYeahs IT Technologies – AI-First Digital Engineering

The client is a sports technology–focused organization exploring AI-driven performance analytics for table tennis at both professional and training levels. They aim to leverage computer vision and deep learning to extract high-speed gameplay insights from video footage. Their objective is to transform raw match videos into structured performance data for coaches, players, analysts, and broadcasters. With a strong interest in innovation, the client seeks scalable, real-time analytics to enhance training effectiveness and viewer engagement.

The client faced challenges due to the lack of granular, frame-level analytics to accurately evaluate table tennis gameplay. Tracking the extremely fast-moving ball proved difficult with conventional video analysis tools. Players and coaches had limited access to objective, data-driven performance metrics. As a result, match analysis relied heavily on manual tagging and annotation, making the process time-consuming and inconsistent. These limitations restricted deeper performance insights and slowed training improvements.

The solution addresses manual and limited video analysis by fully automating frame-level gameplay insights from match footage. It delivers objective, data-driven metrics that support smarter coaching decisions and clearer gameplay understanding. Key innovations include robust ball tracking through combined object detection and segmentation, precise bounce point extraction, and integrated player pose estimation. A modular AI pipeline processes video frames through detection, tracking, trajectory analysis, and visualization stages.

The approach focused on automating table tennis match analysis using advanced computer vision and AI techniques. Full HD videos were ingested and frames extracted for detailed, frame-level processing. Object detection identified players, the table, and the ball, while SAM2 segmentation and trajectory analysis tracked ball movement with precision. Player keypoints and poses were estimated to capture biomechanical insights and movement patterns.

The solution leveraged Python as the core development language, integrating YOLOv8 for player, table, and ball detection, SAM2 for precise ball segmentation and tracking, and MediaPipe for player keypoint and pose estimation. OpenCV handled frame processing and visualization, while NumPy and Pandas managed data structuring and export. Key modules included video ingestion and frame extraction, object detection, ball tracking, and trajectory visualization. Additional modules extracted bounce points, estimated player poses, and exported structured datasets in CSV, Excel, and JSON formats.

The system achieved accurate detection of players across all frames and maintained stable identification of table boundaries. Ball tracking and trajectory mapping were highly reliable, with bounce points successfully extracted for analysis. Player keypoints were consistently detected to capture movement patterns. All data was structured and exported for downstream analytics and performance evaluation. High-quality annotated visuals provided clear, actionable insights from the match footage.

The Table Tennis AI Analysis showcased how computer vision and deep learning can generate valuable insights from match footage. It automates player and ball detection, trajectory mapping, and pose estimation to convert raw video into actionable data. The system provides coaches, analysts, and broadcasters with objective performance metrics. It lays a strong technical foundation for future sports analytics and intelligent coaching applications. It also highlights the potential for AI-enhanced broadcasting and real-time game analysis.

AI-powered solution for actionable table tennis performance insights and competitive advantage

Meet the Client

The Challenges

The Solution

Ready for similar results?

The Approach

Technology and Innovation

The Outcome

Lessons learned