An Improved EfficientNetV2 Model Based on Visual Attention Mechanism: Application to Identification of Cassava Disease

<div>The dynamic change curve of the learning rate of the model drawn by the pseudo-code, the cosine annealing mode is used to decay at 0∼120 epoch, and the value of the learning rate is adjusted to 1/8, 1/16, 1/32 at the 20th, 60th, and 100th epoch.</div>

Computational Intelligence and Neuroscience

fig6

Figure 6

Figure 6: An Improved EfficientNetV2 Model Based on Visual Attention Mechanism: Application to Identification of Cassava Disease 