ABSTRACT
Team "DL_Lock": The Sussex-Huawei Locomotion-Transportation (SHL) recognition challenge 2020 poses a unique opportunity to work on a broad, real-life dataset to classify transport-related activities in a user and location-independent manner. Since deep learning architectures have now received great attention on achieving promising results on time series classification tasks, we focused our experiments on some recent state-of-the-art deep learning architectures such as CNN, Resnet, and InceptionTime. A considerable amount of time was spent on the preprocessing pipeline, which turned out to be a critical phase that impacted most of the results. At the end and after many experiments and hyperparameter tuning, we were able to achieve a 79% F1 score on the validation dataset using InceptionTime architecture. The objective of this paper is to present the technical description of the Machine Learning processing pipeline, the algorithms used, and the results achieved during the development/training phase.
- L. Wang, H. Gjoreski, M. Ciliberto, S. Mekki, S. Valentin, and D. Roggen, "Enabling reproducible research in sensor-based transportation mode recognition with the Sussex-Huawei dataset," IEEE Access 7 (2019): 10870--10891.Google Scholar
- H. Gjoreski, M. Ciliberto, L. Wang, F.J.O. Morales, S. Mekki, S. Valentin, and D. Roggen, "The university of sussex-huawei locomotion and transportation dataset for multimodal analytics with mobile devices," IEEE Access 6 (2018): 42592--42604.Google Scholar
- "Sussex-Huawei Locomotion Challenge.," [Online]. Available: http://www.shl-dataset.org/activity-recognition-challenge-2020/.Google Scholar
- Scikit-learn: Machine Learning in Python, Pedregosa et al., JMLR 12, pp. 2825--2830, 2011.Google Scholar
- Ioffe, Sergey, and Christian Szegedy. "Batch normalization: Accelerating deep network training by reducing internal covariate shift." arXiv preprint arXiv:1502.03167 (2015).Google Scholar
- Wikipedia contributors. (2020, June 26). Euler angles. In Wikipedia, The Free Encyclopedia. Retrieved 08:14, June 29, 2020, from https://en.wikipedia.org/w/index.php?title=Euler_angles&oldid=964664972.Google Scholar
- Barandas, Marília and Folgado, Duarte, et al. "TSFEL: Time Series Feature Extraction Library." SoftwareX 11, 2020. Google ScholarCross Ref
- Wikipedia contributors. (2020, June 11). Local tangent plane coordinates. In Wikipedia, The Free Encyclopedia. Retrieved 08:21, June 29, 2020, from https://en.wikipedia.org/w/index.php?title=Local_tangent_plane_coordinates&oldid=961980629.Google Scholar
- Ronald E. Shiffler (1988) Maximum Z Scores and Outliers, The American Statistician, 42:1, 79--80 Google ScholarCross Ref
- Christ, Maximilian, et al. "Time series feature extraction on basis of scalable hypothesis tests (tsfresh-a python package)." Neurocomputing 307 (2018): 72--77.Google ScholarDigital Library
- Löning, Markus, et al. "sktime: A unified interface for machine learning with time series." arXiv preprint arXiv:1909.07872 (2019).Google Scholar
- Dafne van Kuppevelt, Christiaan Meijer, Vincent van Hees, Mateusz Kuzak "mcfly" (2017). Google ScholarCross Ref
- S. Zagoruyko and N. Komodakis, "Wide Residual Networks," arXiv: Computer Vision and Pattern Recognition, vol., no., p., 2016.Google Scholar
- Fawaz, Hassan Ismail, et al. "Inceptiontime: Finding alexnet for time series classification." arXiv preprint arXiv:1909.04939 (2019).Google Scholar
- Szegedy C, Liu W, Jia Y, Sermanet P, Reed S, Anguelov D, Erhan D, Vanhoucke V, Rabinovich A (2015) Going deeper with convolutions. In: Proceedings of the IEEE conference on computer vision and pattern recognition, pp 1--9.Google ScholarCross Ref
- L. Wang, H. Gjoreski, M. Ciliberto, P. Lago, K. Murao, T. Okita, and D. Roggen. "Summary of the Sussex-Huawei locomotion-transportation recognition challenge 2020", Proceedings of the 2020 ACM International Joint Conference and 2020 International Symposium on Pervasive and Ubiquitous Computing and Wearable Computers, 2020.Google ScholarDigital Library
Index Terms
- Activity recognition for locomotion and transportation dataset using deep learning
Recommendations
Deep learning for sensor-based activity recognition: A survey
Highlights- We survey deep learning based HAR in sensor modality, deep model, and application.
AbstractSensor-based activity recognition seeks the profound high-level knowledge about human activities from multitudes of low-level sensor readings. Conventional pattern recognition approaches have made tremendous progress in the past years. ...
Deep appearance and motion learning for egocentric activity recognition
Egocentric activity recognition has recently generated great popularity in computer vision due to its widespread applications in egocentric video analysis. However, it poses new challenges comparing to the conventional third-person activity recognition ...
Deep Learning for Sensor-based Human Activity Recognition: Overview, Challenges, and Opportunities
The vast proliferation of sensor devices and Internet of Things enables the applications of sensor-based activity recognition. However, there exist substantial challenges that could influence the performance of the recognition system in practical ...
Comments