How to preprocess char-CNN-RNN or skip-thought text embeddings for birds and flowers

hanzhanggit / StackGAN

MIT License

1.86k stars 457 forks source link

How to preprocess char-CNN-RNN or skip-thought text embeddings for birds and flowers #37

Open SpadesQ opened 6 years ago

Lotayou commented 6 years ago

I've been struggling with this too. The original char-CNN-RNN text encoder is implemented in Torch so I guess transporting it to TensorFlow could be tough. However, it's straightforward to load a Torch model in PyTorch with load_lua() function.

Personally, I suggest you use the PyTorch version. The TensorFlow version is crap AFAIC.

SpadesQ commented 6 years ago

@Lotayou how to get skip-thought-embeddings.pickle

rchavezj commented 6 years ago

Is there any dataset other than birds and flowers?

pranman11 commented 6 years ago

@rchavezj Any luck on the datasets other than those?

liwen-tj commented 5 years ago

Anyone find text description of Oxford-102 flower dataset? I download dataset from http://www.robots.ox.ac.uk/~vgg/data/flowers/102/. But failed to find any text info...