FC and RMAC extracted feature vector size

willard-yuan / cnn-cbir-benchmark

CNN CBIR benchmark (ongoing)

193 stars 53 forks source link

FC and RMAC extracted feature vector size #10

Closed spiralhead7 closed 5 years ago

spiralhead7 commented 6 years ago

Can you please let me know the length of feature extracted for a single image of Oxford dataset.(for both fc and rmac). Also will rmac be better than fc in terms of mAP.

willard-yuan commented 5 years ago

The dimension of the extracted feature is 512. The paper of RMAC shows the result for instance retrieval task is better than fc.