Figure 3
From: Neural networks for increased accuracy of allergenic pollen monitoring
Flowchart showing the pollen image classification process. (a) Reference pollen image capture using the custom Fiji macro Pollen_Projector. (b) Images were inserted into a fixed frame and split into training and testing sets. The training set was used for cross-validation and data augmentation (flip, brightness) so as to train the CNNs VGG16, MobileNetv1 and MobileNetv2. Results from the models trained from scratch are compared to results from transfer learning on pre-trained models. (c) Images from before unseen unknown Urticaceae pollen grains are fed to the resulting classifier. Created using https://app.diagrams.net/.
