Fig. 5
From: Classifying polish in use-wear analysis with convolutional neural networks
An example of saliency maps generated for an image (224 × 224 pixels) of bone polish acquired with a 10x objective, 1000 strokes, and 16 patches, classified using a custom CNN (a) and ResNet50 CNN (b). The position of the polish is outlined in the original image (a, top left) for reference. Colder colors represent areas with lower activation, indicating regions that the model focuses less on, while warmer colors highlight areas with higher activation, showing regions the model pays more attention to.
