Open cxj273 opened 9 years ago
Hi, I never thought someone else was going to see this repository, so all the code is synced with my own developing environment. Sorry about the inconvenient.
The reasons could occur by two reasons, first, the code uses CNN related module, that you can find at
https://github.com/jazzsaxmafia/taeksoo/blob/master/cnn_util.py
This is what, in my code, "from taeksoo.cnn_util import *", imports.
The second reason could be the paths in the code. You have to change the paths for caffe, Flickr data, and so on.
Also, the comment around vocab dictionary does not have any specific meaning. Please uncomment it and try.
If you still have problems, let me know.
Thank you
Hi! I am using this repository since the original repository does not contain code for preparing data. Unfortunately, I am getting "KeyError: 'A'" error in the prepare_data function of flickr30k.py file. for cc in caps: 14 seqs.append([worddict[w.lower()] if (w.lower() in worddict and worddict[w.lower()] < n_words) else 1 for w in cc[0].split()]) ---> 15 seqs.append([worddict[w] if worddict[w] < n_words else 1 for w in cc[0].split()]) 16 feat_list.append(features[cc[1]]) 17 KeyError: 'A' Did you face the same problem?
Thanks!
Hi, i actually did not finish implementing this project, but instead did it using Tensorflow https://github.com/jazzsaxmafia/show_attend_and_tell.tensorflow
THe link above is working. You might want to check that one out.
-Taeksoo
Hi, there make_flickr_dataset.py In this script, all the data are used as training data. The generation of dictionary is commented (why?). Could you please let me know how this generation script works for you? Thanks.