Commented section in feature_extract.py

[goncalomcorreia]

Could you update the project with the code on how to get the features from the Caffe VGGNet-19 model? In preprocess_data.py, it reads:

train_feat_file = 'train2014_fixed_fc7_feat.pkl'
val_feat_file = 'val2014_fixed_fc7_feat.pkl'

but there's no explanation on how to get these features. Could you elaborate? I guess they were obtained by the Caffe model referenced in feature_extract.py, but the commented section is incomplete.

[xiaomengyc]

I found this problem too. The output of VGG features are very sparse, but the author published data here is apparently different from that. I have been in this problem for days, hope the author can publish these codes or give some tips!

[zcyang]

The features are extracted with caffe, with the script:
https://github.com/zcyang/imageqa-san/blob/master/data_vqa/feature_extract.py

Since the features are sparse, they are stored in sparse matrix format:
https://github.com/zcyang/imageqa-san/blob/master/src/data_provision_att_vqa.py#L74