An evaluation of different input transformation for the classification of skateboarding tricks by means of transfer learning

• Main Authors: Muhammad Amirul, Abdullah, Muhammad Ar Rahim, Ibrahim, Muhammad Nur Aiman, Shapiee, Mohd Azraai, Mohd Razman, Rabiu Muazu, Musa, Noor Azuan, Abu Osman, Muhammad Aizzat, Zakaria, Anwar, P. P. Abdul Majeed
• Format: Conference or Workshop Item
• Language: English
• Published: Springer 2023
• Subjects: GV Recreation Leisure; Q Science (General); TA Engineering (General). Civil engineering (General); TJ Mechanical engineering and machinery
• Online Access: http://umpir.ump.edu.my/id/eprint/39755/1/An%20evaluation%20of%20different%20input%20transformation%20for%20the%20classification%20of%20skateboarding%20.pdf; http://umpir.ump.edu.my/id/eprint/39755/; https://doi.org/10.1007/978-981-99-0297-2_22

This study aims to investigate the effect of different input images, namely raw data (RAW) and Continuous Wavelet Transform (CWT) towards the discriminating of street skateboarding tricks, i.e., Ollie, Kickflip, Shove-it, Nollie and Frontside 180 through a variety of transfer learning with optimised k-Nearest Neighbors (kNN) pipelines. Six amateur skateboarders participated in the study, executed the aforesaid tricks five times per trick on an instrumented skateboard where six time-domain signals were extracted prior it was transformed to RAW and CWT. It was shown from the study that the CWT-InceptionV3-optimised kNN pipeline could attain an average test and validation accuracy of 90%.