A New Frontier for Activity Recognition

Janko, Vito; Reščič, Nina; Mlakar, Miha; Drobnič, Vid; Gams, Matjaž; Slapničar, Gašper; Gjoreski, Martin; Bizjak, Jani; Marinko, Matej; Luštrek, Mitja

doi:10.1145/3267305.3267518

Cited by 28 publications

(8 citation statements)

References 7 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…In SHL 2018, the best performance achieved by DL approaches (F1 93.9%) (Gjoreski et al, 2018b) is only 1.5 percentage points higher than the best one by ML approaches (92.4%) (Janko et al, 2018). In SHL 2019, the best performance by DL (75.9%) (Choi and Lee, 2019) is 2.5 percentage points lower than the best performance by ML (78.4%) (Janko et al, 2019).…”

Section: Machine Learning Algorithmsmentioning

confidence: 93%

“…Among these classifiers, RF is the most popular one (8 submissions), followed by XGBoost (6 submissions) and MLP (4 submissions). For the recognition tasks, XGBoost (Janko et al, 2018; Kalabakov (Janko et al, 2019) performs the best in SHL 2019 (see Table 4).…”

Section: Machine Learning Algorithmsmentioning

confidence: 99%

“…Ensemble method is a effective approach to tackle the overfitting problem, e.g. by using RF (Antar et al, 2018;Matsuyama et al, 2018;Nakamura et al, 2018;Siraj and et al, 2020;Brajesh and Ray, 2020), XGBoost (Janko et al, 2018;Lu et al, 2019;Kalabakov et al, 2020;Tseng et al, 2020) or ensembles of classifiers (Gjoreski et al, 2018b;Janko et al, 2019;Ahmed et al, 2019;Widhalm et al, 2019). In ML in general, it has been shown numerous times that ensembles with a large number of base predictors (above 50) are less prone to over-fitting.…”

Section: Tackling Over-fittingmentioning

confidence: 99%

See 2 more Smart Citations

Three-Year Review of the 2018–2020 SHL Challenge on Transportation and Locomotion Mode Recognition From Mobile Sensors

Wang

Gjoreski

Ciliberto

et al. 2021

Front. Comput. Sci.

View full text Add to dashboard Cite

The Sussex-Huawei Locomotion-Transportation (SHL) Recognition Challenges aim to advance and capture the state-of-the-art in locomotion and transportation mode recognition from smartphone motion (inertial) sensors. The goal of this series of machine learning and data science challenges was to recognize eight locomotion and transportation activities (Still, Walk, Run, Bus, Car, Train, Subway). The three challenges focused on time-independent (SHL 2018), position-independent (SHL 2019) and user-independent (SHL 2020) evaluations, respectively. Overall, we received 48 submissions (out of 93 teams who registered interest) involving 201 scientists over the three years. The survey captures the state-of-the-art through a meta-analysis of the contributions to the three challenges, including approaches, recognition performance, computational requirements, software tools and frameworks used. It was shown that state-of-the-art methods can distinguish with relative ease most modes of transportation, although the differentiating between subtly distinct activities, such as rail transport (Train and Subway) and road transport (Bus and Car) still remains challenging. We summarize insightful methods from participants that could be employed to address practical challenges of transportation mode recognition, for instance, to tackle over-fitting, to employ robust representations, to exploit data augmentation, and to exploit smart post-processing techniques to improve performance. Finally, we present baseline results to compare the three challenges with a unified recognition pipeline and decision window length.

show abstract

Section: Machine Learning Algorithmsmentioning

confidence: 93%

Section: Machine Learning Algorithmsmentioning

confidence: 99%

Section: Tackling Over-fittingmentioning

confidence: 99%

See 1 more Smart Citation

Three-Year Review of the 2018–2020 SHL Challenge on Transportation and Locomotion Mode Recognition From Mobile Sensors

Wang

Gjoreski

Ciliberto

et al. 2021

Front. Comput. Sci.

View full text Add to dashboard Cite

show abstract

“…Traditional machine learning methods have exhibited significant efficacy in tackling the problem of transportation mode detection, frequently resulting in a high level of accuracy. For example, efficient algorithms, e.g., XGBoost [9], [12], [13] and Random Forest [10], [16], [19], [20], have demonstrated their effectiveness in analyzing data obtained from accelerometers, gyroscopes, magnetometers, linear accelerometers, gravity, orientation, and ambient pressure sensors. Support Vector Machines (SVMs) are frequently utilized in this particular domain [21] and are frequently regarded as a standard against which deep learning methods are compared.…”

Section: B Traditional Machine Learning Methodsmentioning

confidence: 99%

“…Moreover, instead of using raw sensory data, traditional machine learning techniques are often applied for feature transformation. The Fast Fourier Transformation is frequently used for gleaning frequency domain information from sensory data [9], [10], [13], especially the time domain features. For example, Janko et al [9] highlighted the role of expert knowledge in selecting frequency domain features.…”

Section: B Traditional Machine Learning Methodsmentioning

confidence: 99%

Deep CNN-BiLSTM Model for Transportation Mode Detection Using Smartphone Accelerometer and Magnetometer

Tang

Jahan

Roth

2022

2022 IEEE Intelligent Vehicles Symposium (IV)

View full text Add to dashboard Cite

Transportation mode detection from smartphone data is investigated as a relevant problem in the multi-modal transportation systems context. Neural networks are chosen as a timely and viable solution. The goal of this paper is to solve such a problem with a combination model of Convolutional Neural Network (CNN) and Bidirectional-Long short-term memory (BiLSTM) only processing accelerometer and magnetometer data. The performance in terms of accuracy and F1 score on the Sussex-Huawei Locomotion-Transportation (SHL) challenge 2018 dataset is comparable to methods that require the processing of a wider range of sensors. The uniqueness of our work is the light architecture requiring less computational resources for training and consequently a shorter inference time.

show abstract

Adversarial Learning in Accelerometer Based Transportation and Locomotion Mode Recognition

Günthermann

Wang

Simpson

et al. 2022

Intelligent Systems Reference Library

View full text Add to dashboard Cite

A New Frontier for Activity Recognition

Cited by 28 publications

References 7 publications

Three-Year Review of the 2018–2020 SHL Challenge on Transportation and Locomotion Mode Recognition From Mobile Sensors

Three-Year Review of the 2018–2020 SHL Challenge on Transportation and Locomotion Mode Recognition From Mobile Sensors

Deep CNN-BiLSTM Model for Transportation Mode Detection Using Smartphone Accelerometer and Magnetometer

Adversarial Learning in Accelerometer Based Transportation and Locomotion Mode Recognition

Contact Info

Product

Resources

About