Skip to main content

Table 3 The continuous five times classification results

From: Classification methods of a small sample target object in the sky based on the higher layer visualizing feature and transfer learning deep networks

Experimental numerical order (no.)

Accuracy (%)

Non-transfer learning

Transfer learning

1

96.667

96.667

2

90

90

3

93.33

93.33

4

96.667

93.33

5

80

80

Average

91.33

90.67

  1. The total numbers of target object images are 150 pictures, and the sample numbers are all 50 pictures in each type of UAVs, birds, and kites. The training set has 120 pictures of target object images, and the test set has 30 pictures of target object images