Skip to main content

Table 5 Summary of data annotation and partitioning for the TAMU2015, UGA2015, and UGA2018 datasets

From: DeepSeedling: deep convolutional network and Kalman filter for plant seedling detection and counting in the field

Dataset

TAMU2015

UGA2015

UGA2018

Total number of images

2204

1895

1511

Number of training images

1801

1603

1253

Number of validation images

202

146

129

Number of testing images

201

146

129

Total number of annotations (plant seedling/weed)

21915 (21133/782)

7802 (6849/953)

5880 (5880/0)

Number of training annotations (plant seedling/weed)

17939 (17290/649)

6524 (5743/781)

4862 (4862/0)

Number of validation annotations (plant seedling/weed)

1964 (1911/53)

643 (553/90)

540 (540/0)

Number of testing annotations (plant seedling/weed)

2012 (1932/80)

635 (553/82)

478 (478/0)

Number of videos for counting test

25

25

25