Page 181 - Handbook of Deep Learning in Biomedical Engineering Techniques and Applications

170 Chapter 6 Plant leaf disease classification based on feature selection

(A)
Training and Validation accuracy
90%
80%
70%
60%
50%
40%
30%
20%
10%
0%
123456789101112131415161718192021222324252627282930
Training acc Validation acc
(B)

Figure 6.14 Training and validation result of ResNet-18 model.

Table 6.3 Confusion matrix of ResNet-18 model.

Class C1 C2 C3 C4
C1 16 6 3 2
C2 5 17 1 0
C3 0 0 26 0
C4 3 0 0 6

C1, anthracnose; C2, gall midge; C3, healthy; C4, powdery mildew. ResNet-18 achieved
84.1% training accuracy and 76.5% testing accuracy.