Skip to main content

Table 5 Average performance of the baseline multi-class CNN and best trained models (CNN-LSTM) on test data before and after post-processing step

From: Deep learning-based detection of seedling development

Model

Accuracy

Error

Sensitivity

Specificity

Precision

False positive rate

Multi-class CNN (Before)

0.72 ± 0.29

0.28 ± 0.29

0.73  ± 0.19

0.94 ± 0.21

0.91 ± 0.13

0.8 ± 0.08

Multi-class CNN (After)

00.80 ± 0.19

0.20 ± 0.19

0.85 ± 0.13

0.93 ± 0.07

0.85 ± 0.14

0.07 ± 0.07

CNN-LSTM (Before)

0.84 ± 0.04

0.16 ± 0.04

0.83  ± 0.05

0.93 ± 0.06

0.86 ± 0.09

0.05 ± 0.05

CNN-LSTM (After)

0.90 ± 0.08

0.10 ± 0.07

0.87 ± 0.11

0.96 ± 0.03

0.88 ± 0.15

0.04 ± 0.04