Skip to main content

Table 3 Attention score statistics of multi-scale detection layers

From: Improving multi-scale detection layers in the deep learning network for wheat spike detection based on interpretive analysis

Detection layer

Attention score

Proportion of spikes (%)

Average spike size (Pixels)

Total attention score range

Mean attention score

Small-scale detection layer

0.0–0.1

64.4

8682

0–0.607

0.083

0.1–0.2

27.5

3438

0.2–0.3

6.3

2137

0.3–0.4

1.4

1581

0.4–0.5

0.3

1188

0.5–0.6

0.07

961

0.6–0.7

0.03

733

0.7–0.8

–

–

0.8–0.9

–

–

0.9–1.0

–

–

Medium-scale detection layer

0.0–0.1

13.0

12030

0–0.984

0.25

0.1–0.2

26.4

8949

0.2–0.3

27.9

5619

0.3–0.4

17.8

4165

0.4–0.5

9.0

3360

0.5–0.6

3.9

2758

0.6–0.7

1.4

2347

0.7–0.8

0.5

2003

0.8–0.9

0.08

1742

0.9–1.0

0.02

1527

Large-scale detection layer

0.0–0.1

38.8

3891

0–1

0.309

0.1–0.2

6.3

10298

0.2–0.3

7.3

11365

0.3–0.4

8.7

10581

0.4–0.5

9.0

8946

0.5–0.6

9.2

7962

0.6–0.7

8.2

6861

0.7–0.8

6.6

6092

0.8–0.9

4.0

5297

0.9–1.0

1.9

4346