search

chenbinghui1 / MHN

ICCV2019
183 stars 35 forks source link

Detail information about "fea" and "y" #7

Closed liaojiacai closed 4 years ago

liaojiacai commented 4 years ago

Dear author: Thanks for your great work, your paper and code are perfect. A good idea has been proposed in your paper, and clearly code realized it, what's more, your idea has a strong theoretical persuasion. While, when I print the output of the network, it was hard for me to understand the size of for an input image, parts=4, part = 3, the output size of "fea" are: net output size"fea": 3x4x256 can I understand that for every input image, the networks will generate 12 features, and every feature's dimension is 256?

chenbinghui1 commented 4 years ago

@liaojiacai Yes,you are right!