Assume I have a tensor of shape (b,h,w,d). It is:
b: batch_size
h: height of image
w: width of image
d: feature dimension
How would I sample a set of 100 pixels for each pixel of it, such that I have a tensor of shape (b, 100, d)?
Also, how would I do it if do not know h, w, which is the cast at test time?
Aucun commentaire:
Enregistrer un commentaire