Hello @lahammond & collaborators,
As mentioned in a previous issue, I am trying to implement RESPAN for spine analysis over time, meaning I am looking for newly formed spines - aka filopodia. Model1B is the most sensitive model for spines on our spinning-disk images, however it is weak for filopodia detection. During the Model1B training on the 44 datasets, I guess there was a quantity of filopodia. I'm curious to ask how you annotated filopodia during the training? As a spine neck only? Did you simply not consider filopodia as spines?
Thank you in advance,
Loïc
Hello @lahammond & collaborators,
As mentioned in a previous issue, I am trying to implement RESPAN for spine analysis over time, meaning I am looking for newly formed spines - aka filopodia. Model1B is the most sensitive model for spines on our spinning-disk images, however it is weak for filopodia detection. During the Model1B training on the 44 datasets, I guess there was a quantity of filopodia. I'm curious to ask how you annotated filopodia during the training? As a spine neck only? Did you simply not consider filopodia as spines?
Thank you in advance,
Loïc