TY - GEN
T1 - The Effect of Image Input Transformation from Inertial Measurement Unit Data on the Classification of Skateboarding Tricks
AU - Abdullah, Muhammad Amirul
AU - Ibrahim, Muhammad Ar Rahim
AU - Shapiee, Muhammad Nur Aiman
AU - Zakaria, Muhammad Aizzat
AU - Mohd Razman, Mohd Azraai
AU - Musa, Rabiu Muazu
AU - P. P. Abdul Majeed, Anwar
N1 - Funding Information:
Acknowledgement. The authors would like to thank the Ministry of Higher Education (MoHE) for funding the study through the Fundamental Research Grant Scheme viz. FRGS/1/2019/TK03/UMP/02/6 (RDU1901115).
Publisher Copyright:
© 2021, The Author(s), under exclusive license to Springer Nature Singapore Pte Ltd.
PY - 2021
Y1 - 2021
N2 - This study aims to improve the classification of flat-ground tricks namely Ollie, Kickflip, Shove-it, Nollie and Frontside 180 through the identification of significant input image transformation on different transfer learning models. Six goofy skateboarders (23 years of age ± 5.0 years’ experience) executed five tricks for each type of trick repeatedly on a customized ORY skateboard (Inertial Measurement Unit (IMU) sensor fused) on a cemented ground. From the IMU data, six raw signals were extracted. The best input image transformation and transfer learning model were identified through two input image transformations synthesized, namely raw data (RAW) and Fast Fourier Transform (FFT), and six transfer learning models based on default arguments from the Keras library. The variation of the SVM models (via different hyperparameters) was evaluated both on input image transformation and on transfer learning model in classifying the skateboarding tricks. It was shown from the study that RAW input image on three transfer learning models (DenseNet121, InceptionResNetV2, and ResNet101) demonstrated the 100% accuracy on all train, train and validation dataset. It could be concluded that the proposed method is able to improve the classification of the skateboarding tricks well.
AB - This study aims to improve the classification of flat-ground tricks namely Ollie, Kickflip, Shove-it, Nollie and Frontside 180 through the identification of significant input image transformation on different transfer learning models. Six goofy skateboarders (23 years of age ± 5.0 years’ experience) executed five tricks for each type of trick repeatedly on a customized ORY skateboard (Inertial Measurement Unit (IMU) sensor fused) on a cemented ground. From the IMU data, six raw signals were extracted. The best input image transformation and transfer learning model were identified through two input image transformations synthesized, namely raw data (RAW) and Fast Fourier Transform (FFT), and six transfer learning models based on default arguments from the Keras library. The variation of the SVM models (via different hyperparameters) was evaluated both on input image transformation and on transfer learning model in classifying the skateboarding tricks. It was shown from the study that RAW input image on three transfer learning models (DenseNet121, InceptionResNetV2, and ResNet101) demonstrated the 100% accuracy on all train, train and validation dataset. It could be concluded that the proposed method is able to improve the classification of the skateboarding tricks well.
KW - Classification
KW - Machine learning
KW - Skateboarding
KW - Support Vector Machine
KW - Transfer learning
UR - http://www.scopus.com/inward/record.url?scp=85113816756&partnerID=8YFLogxK
U2 - 10.1007/978-981-16-4803-8_42
DO - 10.1007/978-981-16-4803-8_42
M3 - Conference Proceeding
AN - SCOPUS:85113816756
SN - 9789811648021
T3 - Lecture Notes in Mechanical Engineering
SP - 424
EP - 432
BT - RiTA 2020 - Proceedings of the 8th International Conference on Robot Intelligence Technology and Applications
A2 - Chew, Esyin
A2 - P. P. Abdul Majeed, Anwar
A2 - Liu, Pengcheng
A2 - Platts, Jon
A2 - Myung, Hyun
A2 - Kim, Junmo
A2 - Kim, Jong-Hwan
PB - Springer Science and Business Media Deutschland GmbH
T2 - 8th International Conference on Robot Intelligence Technology and Applications, RiTA 2020
Y2 - 11 December 2020 through 13 December 2020
ER -