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_52
Snippet: The models are implemented in PyTorch. Batch normalization [55] is used through all models. Binary cross-entropy serves as the loss function. The networks are trained with four GTX 1080Ti GPUs using data parallelism. Hyperparameters are tuned on the validation set. Data augmentation is implemented to further improve generalization. For each image in our COVID19-CT dataset, we apply different random affine transformations including random cropping.....
Document: The models are implemented in PyTorch. Batch normalization [55] is used through all models. Binary cross-entropy serves as the loss function. The networks are trained with four GTX 1080Ti GPUs using data parallelism. Hyperparameters are tuned on the validation set. Data augmentation is implemented to further improve generalization. For each image in our COVID19-CT dataset, we apply different random affine transformations including random cropping with a scale of 0.5 and horizontal flip. Color jittering is also applied with random contrast and random brightness with a factor of 0.2.
Search related documents:
Co phrase search for related documents- horizontal flip and random brightness: 1
- horizontal flip and random contrast: 1
- horizontal flip and random cropping: 1
- loss function and PyTorch implement: 1
- loss function and validation set: 1
Co phrase search for related documents, hyperlinks ordered by date