Enhancing Human Activity Recognition through Transformer-based GANs

Ihab AHMED

Diplomová práce

Abstract:

A custom Transformer-based GAN tailored for time-series data was developed to improve Human Activity Recognition systems. The effectiveness of the synthetic data generated was assessed, showing dependence on the diversity and quality of the training datasets. The synthetic data's impact on enhancing system performance and its potential for privacy-sensitive applications were also explored.

Abstract:

A custom Transformer-based GAN tailored for time-series data was developed to improve Human Activity Recognition systems. The effectiveness of the synthetic data generated was assessed, showing dependence on the diversity and quality of the training datasets. The synthetic data's impact on enhancing system performance and its potential for privacy-sensitive applications were also explored.

Keywords

Deep Learning, Artificial Neural Networks, Generative AI, Transformers, Generative Adverserial Networks, time-series data analysis, Human Activity Recognition

Jazyk práce: angličtina

Datum vytvoření / odevzdání či podání práce: 20. 8. 2024

Obhajoba závěrečné práce

• Vedoucí: prof. Dr. Patrick Glauner

JIHOČESKÁ UNIVERZITA V ČESKÝCH BUDĚJOVICÍCH

Přírodovědecká fakulta

Magisterský studijní program / obor:
Artificial Intelligence and Data Science / Artificial Intelligence and Data Science