MTAL: A Novel Chinese Herbal Medicine Classification Approach with Mutual Triplet Attention Learning

Figure 4

Comparison results of different models in terms of accuracy and loss based on the different basic student networks. (a) Comparison of accuracy of different models, including single ResNet18 model (SigRes18), single ResNet18+triplet attention (SigAttRes18), single ResNet18 with knowledge learned from ResNet50 (MulRes18(50)), SigAttRes18 with knowledge learned from SigAttRes50 (MTALRes18(50)). (b) Comparison of accuracy of different models, including single ResNet50 model (SigRes50), single ResNet50+triplet attention(SigAttRes50), single ResNet50 with knowledge learned from ResNet18 (MulRes50(18)), and SigAttRes50 with knowledge learned from SigAttRes18 (MTALRes50(18)). (c) The comparison of loss of the corresponding models of (a). (d) The comparison of loss of the corresponding models of (b).

(a)

(b)

(c)

(d)