Feature extraction from CRF-RNN vs FCN8s

torrvision / crfasrnn

This repository contains the source code for the semantic image segmentation method described in the ICCV 2015 paper: Conditional Random Fields as Recurrent Neural Networks. http://crfasrnn.torr.vision/

Other

1.34k stars 462 forks source link

Feature extraction from CRF-RNN vs FCN8s #143

Open AlexTS1980 opened 7 years ago

AlexTS1980 commented 7 years ago

I've been extracting features from CRF-RNN and FCN8s, and I noticed one weird thing: When I extract features:


        feats = net.forward(end='fc7', **{'data': caffe_in}) 
        feats = feats['fc7'][0]

Features from FCN8s are twice the size of CRF-RNN. Why would that be the case?