Guoqing Bao

Guoqing Bao

Senior staff researcher at Enflame Technology, obtained PhD in Artificial Intelligence & Image Processing from School of Computer Science, University of Sydney. Currently doing research in AI framework optimization & computer vision. Before his oversea study in Australia, he had worked as a senior software engineer in Beijing (e.g., iQiYi.com) from 2008 to 2016.

COVID-MTL: Multitask Learning with Shift3D and Random-weighted Loss for COVID-19 Diagnosis and Severity Assessment

Published in Pattern Recognition, 2021

Recommended citation: Guoqing Bao et al., (2021). "COVID-MTL: Multitask Learning with Shift3D and Random-weighted Loss for COVID-19 Diagnosis and Severity Assessment" Pattern Recognition, 2021, doi: 10.1016/j.patcog.2021.108499. https://doi.org/10.1016/j.patcog.2021.108499

We proposed a multitask learning framework for COVID-19 diagnosis and serverity assessment. The Shift3D algorithms improved the performance of 3D CNN and the random-weighted loss addressed the problem of task dominance in joint learning various imbalanced tasks.

Corresponding full-text paper from Pattern Recognition: https://www.sciencedirect.com/science/article/pii/S0031320321006750

Open-source Code in GitHub: link.