Fig. 8: Effect of the number of micrographs used for training.

Graph showing the average precision values for coordinates of TcdA1 and KLH which are predicted by models trained using different numbers of micrographs. It is to be noted that the number varies for different datasets. Training with at least 12 micrographs was required for KLH, whereas training with 20 micrographs was required for TcdA1 to get a prediction AUC score above 0.9.