Selected article for: "image transfer and large scale"

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_17
    Snippet: Transfer learning is normally performed by taking a standard neural architecture along with its pretrained weights on large-scale datasets such as ImageNet [20] , and then finetuning the weights on the target task. This idea has been successfully applied to visual recognition [21] as well as language comprehension [22] . In the medical domain, transfer learning has also been widely used in medical image classification and recognition tasks, such .....
    Document: Transfer learning is normally performed by taking a standard neural architecture along with its pretrained weights on large-scale datasets such as ImageNet [20] , and then finetuning the weights on the target task. This idea has been successfully applied to visual recognition [21] as well as language comprehension [22] . In the medical domain, transfer learning has also been widely used in medical image classification and recognition tasks, such as tumor classification [23] , retinal diseases diagnosis [24] , pneumonia detection [25] , and skin lesion and cancer classification [26] , [27] . A recent study in [28] explores the properties of transfer learning for medical imaging tasks and finds that the standard large networks pretrained on ImageNet are often over-parameterized and may not be the optimal solution for medical image diagnosis. In this paper, we continue to investigate different strategies of transfer learning and integrate contrastive self-supervised learning into the transfer learning process to learn powerful and unbiased feature representations for reducing the risk of overfitting.

    Search related documents:
    Co phrase search for related documents
    • cancer classification and medical domain: 1, 2
    • cancer classification and medical image: 1, 2, 3, 4
    • cancer classification and medical image classification: 1, 2
    • cancer classification and medical imaging: 1
    • different strategy and large scale: 1, 2, 3, 4, 5
    • different strategy and medical imaging: 1, 2
    • different strategy and optimal solution: 1, 2, 3, 4
    • feature representation and large scale: 1, 2, 3
    • feature representation and large scale dataset: 1
    • feature representation and medical image: 1, 2
    • feature representation and medical imaging: 1, 2, 3
    • feature representation and neural architecture: 1
    • large network and medical domain: 1, 2
    • large network and medical image: 1, 2, 3
    • large network and medical image classification: 1, 2
    • large network and medical imaging: 1, 2, 3, 4, 5, 6, 7
    • large network and neural architecture: 1, 2, 3, 4
    • large network and optimal solution: 1, 2
    • large scale and medical domain: 1, 2, 3, 4, 5, 6, 7, 8, 9, 10, 11, 12, 13, 14, 15