Selected article for: "deep learning and transfer learning"

Author: Xuehai He; Xingyi Yang; Shanghang Zhang; Jinyu Zhao; Yichen Zhang; Eric Xing; Pengtao Xie
Title: Sample-Efficient Deep Learning for COVID-19 Diagnosis Based on CT Scans
  • Document date: 2020_4_17
  • ID: l3f469ht_74
    Snippet: In this paper, we study how to develop sample-efficient deep learning methods to accurately diagnose COVID-19 from CT scans. To facilitate the open research in this area, we build COVID19-CT, a dataset containing 349 CT scans positive for COVID-19. To our best knowledge, it is the largest COVID19-CT dataset that is publicly available to date. Though the largest, it still incurs a high risk of overfitting for data-hungry deep learning models. To r.....
    Document: In this paper, we study how to develop sample-efficient deep learning methods to accurately diagnose COVID-19 from CT scans. To facilitate the open research in this area, we build COVID19-CT, a dataset containing 349 CT scans positive for COVID-19. To our best knowledge, it is the largest COVID19-CT dataset that is publicly available to date. Though the largest, it still incurs a high risk of overfitting for data-hungry deep learning models. To reduce this risk, we develop data efficient methods that are able to mitigate data deficiency. We propose Self-Trans, a self-supervised transfer learning approach that learns expressive and unbiased visual feature representations that are robust to overfitting. Through extensive experiments, we demonstrate the effectiveness of our methods. is the (which was not peer-reviewed) The copyright holder for this preprint . https://doi.org/10.1101/2020.04.13.20063941 doi: medRxiv preprint

    Search related documents:
    Co phrase search for related documents