Laser & Optoelectronics Progress, Volume. 60, Issue 4, 0410004(2023)
Fine-Grained Classification of Wild Mushrooms Based on Feature Fusion and Attention Mechanism
Figures & Tables(14)
Fig. 8. Experimental process of different models on the validation set. (a) Accuracy convergence curve; (b) loss convergence curve
Fig. 9. Experimental process of the comparison experiment on the validation set. (a) Accuracy convergence curve; (b) loss convergence curve
Fig. 11. Experimental process of MobileNet_v2 combined with PA_CBAM on the validation set. (a) Accuracy convergence curve; (b) loss convergence curve
Table 1. Recognition results of different models
View tableTable 1. Recognition results of different models
Model Accuracy /% Precision /% Recall /% F1 score /% AlexNet 75.43 76.58 75.53 75.78 Vgg19 76.55 76.87 76.65 76.74 SqueezeNet 71.68 71.81 71.71 71.44 ShuffleNet_v2 76.37 76.37 76.37 76.21 MobileNet_v2 77.59 77.66 77.60 77.47 Inception_v3 82.43 82.66 82.43 82.38 DenseNet121 84.81 85.39 85.10 85.08 ResNet50 85.17 85.63 85.13 85.40 I_ResNet50 86.03 86.00 85.97 86.24
Table 2. Comparative experiment results
View tableTable 2. Comparative experiment results
Model Accuracy of Top1 /% Accuracy of Top5 /% Loss Recall /% Precision /% ResNet50 85.17 96.46 0.5552 85.13 85.15 I_ResNet50 86.03 97.19 0.5212 85.97 86.00 ResNet50+channel 84.85 95.81 0.6178 84.85 84.85 ResNet50+spatial 83.71 96.89 0.5643 83.62 83.81 ResNet50+CBAM 84.74 96.39 0.6214 84.55 84.68 I_ResNet50+CBAM 85.49 96.89 0.5797 85.49 85.62 ResNet50+PA_CBAM 87.66 97.11 0.5335 87.14 87.47 I_ResNet50+PA_CBAM 88.52 97.58 0.4927 88.53 88.71 Model F1 score /% Predicting time /ms Size /MB Parameter number /107 Training time /h ResNet50 85.40 1.557 270.06 2.332 12.41 I_ResNet50 86.24 1.629 525.81 4.572 11.83 ResNet50+channel 84.48 1.592 286.50 2.356 4.50 ResNet50+spatial 83.60 1.410 283.20 2.344 4.00 ResNet50+CBAM 84.48 2.189 298.92 2.607 5.25 I_ResNet50+CBAM 85.37 3.496 554.66 4.824 9.67 ResNet50+PA_CBAM 87.71 2.259 298.92 2.807 5.42 I_ResNet50+PA_CBAM 88.48 3.365 554.66 4.824 9.33
Table 3. PA_CBAM versatility verification experiment results
View tableTable 3. PA_CBAM versatility verification experiment results
Model Accuracy /% Predicting time /ms Size /MB AlexNet + PA_CBAM 76.91 Vgg19 + PA_CBAM - - - MobileNet_v2 + PA_CBAM 79.18 1.27 27.7 MobileNet_v2（pretrained） 94.21 1.65 27.8 MobileNet_v2（pretrained） + PA_CBAM 94.87 1.30 27.8
Tools
Get Citation
Copy Citation Text
Jiaxin Qian, Pengfei Yu, Haiyan Li, Hongsong Li. Fine-Grained Classification of Wild Mushrooms Based on Feature Fusion and Attention Mechanism[J]. Laser & Optoelectronics Progress, 2023, 60(4): 0410004
Paper Information
Category: Image Processing
Received: Oct. 21, 2021
Accepted: Dec. 21, 2021
Published Online: Feb. 14, 2023
The Author Email: Pengfei Yu (pfyu@ynu.edu.cn)
© Copyright 2018-2021 | Chinese Laser Press.
All Rights Reserved 沪ICP备15018463号-20