AAI_2025_Capstone_Chronicles_Combined

12

Figure7 Misclassified image examples, CNN model

Now, looking at the ViT model, it achieved slightly less accuracy than the CNN model with 81.25% correctly classified samples: Table2 Training test results of the ViT model Precision Recall F1Score Fake 0.81 0.82 0.82 Real 0.82 0.80 0.81 On the other hand, we can see that the predictions made by the ViT model are way more balanced than those made by the CNN. In this case, there does not seem to have any bias towards any specific label.

369