Kyfafyd / MirrorGAN

Reproduction of the paper MirrorGAN: Learning Text-to-image Generation by Redescription
MIT License
25 stars 4 forks source link

train the veri776 dataset #11

Open Yuchen-Hou0210 opened 2 years ago

Yuchen-Hou0210 commented 2 years ago

hello, I want to do some experiments on veri776 datasets. I have classified the pictures and the text I wrote into folders by category. But I don't have pickle file and json file, I would like to ask if I can generate these contents after I create these empty files and run preprocess.py? I would like to know what is caption.pickle used for? I opened the caption.pickle of birds and found that it has four parts, which represent train_captions, test_captions, ixtoword, wordtoix. I want to know what the numbers in train_captions or test_captions represent? btw,i think you are Chinese , can I communicate with you in Chinese? it can express more clearly