Automated tool for swimmer analysis

• Main Author: Soh, Jun Feng
• Other Authors: Lam Siew Kei
• Format: Final Year Project
• Language: English
• Published: Nanyang Technological University 2020
• Subjects: Engineering::Computer science and engineering::Computing methodologies::Image processing and computer vision
• Online Access: https://hdl.handle.net/10356/137983
• Institution: Nanyang Technological University

In this paper, an evaluation of existing swimming analysis methodologies was conducted to determine their strengths and weaknesses. Some of the strengths include the ability to maintain a swimmer’s mobility and receiving the guidance of a coach. With the aim to build towards the benefits of existing products, an open source human pose network was adopted to detect human pose based on video frames. Designs were then implemented to identify bad swim strokes from the human pose detected. Suggestions will then be provided to correct the swimmers. This tool was then evaluated to determine the effectiveness. It was concluded that due to inaccuracies with the network, the swim analysis tool did not work as well as intended. However, the concept of implementing a coach’s guidance into the tool is possible, whereas the network requires further evaluation and possibly training on underwater datasets.