Skip to main content

Table 8 Effect on the specular highlight grafting augmentation during pretraining of the encoder with the DTD data set

From: Automated classification of polyps using deep learning architectures and few-shot learning

Model

finetuning

highlight grafting

Acc

Pre

Rec

F1

DTD 

Y

N

80.44

80.51

80.44

80.3

 

Y

Y

81.34

81.52

81.34

81.0

 

N

N

65.81

63.86

65.81

64.7

 

N

Y

68.61

71.92

68.61

70.1

  1. The average performance on 100 random train/test splits is reported. Bold values are indicating the highest value of a column