Skip to main content
. 2023 Feb 24;13:3205. doi: 10.1038/s41598-023-30064-6

Figure 2.

Figure 2

(a) Input sample and CNN architecture. (b) Number of samples before and after filtering. (c) Averaged confusion matrix across 10 experimental runs. (d) Variability in accuracy across pollen classes in 10 experimental runs.