Table 1. Previous attempts at automated classification of 10 or more pollen types, with number of taxa, classifier type and highest reported correct classification rate (CCR).
Abbreviations: NN = neural network, SVM = support vector machine, LDA = linear discriminant analysis, KNN = k-nearest neighbour, CNN = convolutional neural network, RNN = recurrent neural network.
| # taxa | classifier | reported CCR | |
|---|---|---|---|
| Daood et al. (2016) [21] | 30 | CNN | 94% |
| Sevillano & Aznarte (2018) [10] | 23 | CNN | 97% |
| Goncalves et al. (2016) [12] | 23 | SVM | 64% |
| Kaya et al. (2013) [22] | 19 | Rough Set | 91% |
| Ticay-Rivas et al. (2011) [23] | 17 | NN | 96% |
| Marcos et al. (2015) [24] | 15 | KNN | 95% |
| Lagerstrom et al. (2015) [25] | 15 | LDA | 81% |
| Li et al. (2004) [26] | 13 | LDA, NN | 100% |
| Treloar et al. (2004) [27] | 12 | LDA | 100% |
| Khanzhina et al. (2018) [19] | 11 | CNN | 96% |
| Dhawale et al. (2013) [28] | 10 | NN | 85% |
| Daood et al. (2018) [18] | 10 | CNN+RNN | 100% |