Golf Club
Highlights
- Detection of positions of player’s hands and golf club with computer vision technology
- Two AI models for detecting and segmenting the images for metrics collection
- Collection of advanced metrics for further analysis and processing
- Robust pose estimation achieved by measuring angles of the given body joints with computer vision
Challenge
The client is a manufacturer of custom golf clubs from New Zealand. When he approached SoftTeco, he had the idea of developing an AI-based product for analyzing the players’ positions and strokes in order to design individually fitting golf clubs.
At that time, the client was already using computer analysis software for analyzing the players’ movements and consistency of strokes. But the work was semi-manual and the client wanted to fully automate the process. Therefore, he contacted SoftTeco as a trusted software provider.
Solution
The product is an AI-powered solution that is capable of recognizing the golf club in a player’s hand and correctly estimating the angles of his joints (the posture) with an aim to collect and process these metrics and transform them into final technical requirements for club manufacturing.
The metrics are captured either via a mobile device camera or via a pre-installed kiosk with a camera. By using the collected data, the client can manufacture individual golf clubs that are adjusted to one’s level of impact, playing posture, and strokes.
Tech Stack
Components
Keras
Python
Matplotlib
Mediapipe
MLFlow
NumPy
TensorFlow
OpenCV
LabelMe
Colab
How it works
The first challenge that the SoftTeco team faced was a lack of data for ML model training. In order to create a suitable dataset, we had to collect videos from available sources and transform them into suitable images. These images were formed in a dataset and later labeled by two categories: images for detection and images for segmentation.
Two AI models were designed correspondingly. One model is responsible for detection: it detects the position of a player’s hand and the position of a club. The second model performs segmentation: it recognizes the club on the given image and differentiates it from the surroundings and the player. These two models work in conjunction and collect all the metrics needed to create an individual and perfectly fitting golf club. The collected metrics include angles of body joints during the movement, position of a player’s body, and their golf club during the stroke. Computer vision technology can recognize a golf club from its surroundings and can differentiate between a player’s arm and a club.
Results
The product was developed as an MVP and performs all the intended functions with the help of computer vision and artificial intelligence technologies. The client remained highly content with the quality and speed of SoftTeco’s work and we hope to continue working on this project in the future.
Need a high-performing solution to augment your business?
Let SoftTeco handle your tech challenges and design a software solution tailored precisely to your unique requirements.
