Roy Alia Asiku
Current Organizations / Institutes
The organizations/institutes from the scholar's latest article
Roy Alia Asiku has a total of 6 co-authors. Here are the top co-authors by number of shared publications:
- Alice Mugisha
Makerere University Kampala Uganda
1 shared publication - Halimu Chongomweru
College of Computing and Information Sciences, Makerere University, Kampala P.O. Box 7062, Uganda
1 shared publication - Hawa Nyende
Makerere University Kampala Uganda
1 shared publication - Ivan Kayongo
University of Trento Trento Italy
1 shared publication - Lillian Muyama
Makerere University Kampala Uganda
1 shared publication - Michael Kizito
Makerere University Kampala Uganda
1 shared publication
Latest Publications
- Open Access
MakOne: Behavioural Data of University Students' Smart Devices in Uganda
Understanding student behaviour in higher education is essential for improving academic performance, supporting mental well-being, and informing institutional policies. However, most existing behavioural datasets originate from Western institutions and overlook the unique socioeconomic and infrastructural contexts of African institutions, limiting the global applicability of resulting insights. This paper introduces MakOne, a novel multimodal dataset collected over six weeks from 72 students at Makerere University, Kampala, using iLog, a mobile sensing application. The dataset integrates passive smartphone sensor data-including location, physical activity, and screen usage-with ecological momentary assessments (EMAs) that capture students' moods and daily routines. Designed to reflect the lived experiences of students in an African setting, MakOne offers a foundation for research in behaviour modeling, inclusive context-aware system design, mental health analytics, and culturally grounded educational technologies. It contributes a critical African perspective to the growing body of data-driven studies on student behaviour.
Open AccessMakOne: Behavioural Data of University Students' Smart Devices in Uganda
Understanding student behaviour in higher education is essential for improving academic performance, supporting mental well-being, and informing institutional policies. However, most existing behavioural datasets originate from Western institutions and overlook the unique socioeconomic and infrastructural contexts of African institutions, limiting the global applicability of resulting insights. This paper introduces MakOne, a novel multimodal dataset collected over six weeks from 72 students at Makerere University, Kampala, using iLog, a mobile sensing application. The dataset integrates passive smartphone sensor data-including location, physical activity, and screen usage-with ecological momentary assessments (EMAs) that capture students' moods and daily routines. Designed to reflect the lived experiences of students in an African setting, MakOne offers a foundation for research in behaviour modeling, inclusive context-aware system design, mental health analytics, and culturally grounded educational technologies. It contributes a critical African perspective to the growing body of data-driven studies on student behaviour.
- Open Access
- 30 July 2023
- Published by Institute of Electrical and Electronics Engineers (IEEE) in TechRxiv
Automatic classification of land cover with Sentinel-1/2 data at high resolution with low-resolution reference data
In remote sensing, automatic large-scale land cover mapping remains one of the most important problems. There is a vast amount of radar and multi-spectral imagery from satellites such as Sentinel-1 and Sentinel-2 of the European Space Agency’s Copernicus satellite constellation program. Despite the abundance of satellite data, high-resolution, accurately labeled training data remains scarce in this field. This is mainly due to the high cost associated with the labeling of such huge amounts of high-resolution data. Recent works have leveraged existing global land cover products to train land cover mapping models. However, these products often have significant label noise and low resolution, limiting the achievable accuracy of the trained models. In solving this problem, a host of weak-supervision methods have been proposed, including the design of efficient loss functions. In this work, using the SEN12MS dataset and the DFC2020 dataset, we evaluate the performance of previously proposed weak supervision loss functions used for training deep learning models for land cover mapping. We also explore the use of ensembles of losses. We enumerate the performance of popular weak-supervision loss functions and the proposed ensembles. A variation of the backward loss correction approach is shown to perform the best in both datasets, followed by the cross-entropy loss and an ensemble loss composed of the cross-entropy, focal loss, and unhinged loss.
- 30 July 2023
- Published by Institute of Electrical and Electronics Engineers (IEEE) in TechRxiv
Open AccessAutomatic classification of land cover with Sentinel-1/2 data at high resolution with low-resolution reference data
In remote sensing, automatic large-scale land cover mapping remains one of the most important problems. There is a vast amount of radar and multi-spectral imagery from satellites such as Sentinel-1 and Sentinel-2 of the European Space Agency’s Copernicus satellite constellation program. Despite the abundance of satellite data, high-resolution, accurately labeled training data remains scarce in this field. This is mainly due to the high cost associated with the labeling of such huge amounts of high-resolution data. Recent works have leveraged existing global land cover products to train land cover mapping models. However, these products often have significant label noise and low resolution, limiting the achievable accuracy of the trained models. In solving this problem, a host of weak-supervision methods have been proposed, including the design of efficient loss functions. In this work, using the SEN12MS dataset and the DFC2020 dataset, we evaluate the performance of previously proposed weak supervision loss functions used for training deep learning models for land cover mapping. We also explore the use of ensembles of losses. We enumerate the performance of popular weak-supervision loss functions and the proposed ensembles. A variation of the backward loss correction approach is shown to perform the best in both datasets, followed by the cross-entropy loss and an ensemble loss composed of the cross-entropy, focal loss, and unhinged loss.