aai-institute / ProtoTree

Fork of ProtoTrees: Neural Prototype Trees for Interpretable Fine-grained Image Recognition, published at CVPR2021
Other
3 stars 1 forks source link

Using distance or similarity metric for each prototype should be configurable #56

Open gpitteri opened 1 year ago

gpitteri commented 1 year ago

In the current implementation the prototype layer of the network returns only the distances between each prototype and the latent space patches. In the original ProtoPNet the distances are then modified into similarities. We can add this feature.

A configurable parameter set by the user will then decide if the network uses the distances or the similarities values.