gcucurull/visual-compatibility

How to use FashionGen dataset

Opened this issue · 2 comments

Hi, Gcucurull:

I want to ask how to use the FashionGen dataset as said in the paper there is also experiments with that.

I tried to download FashionGen dataset from https://fashion-gen.com/dataset, then read the fashiongen_256_256_train.h5. I only get the keys of metadata like these:

>>> file_h5 = h5py.File("fashiongen_256_256_train.h5", mode="r")
>>> print(file_h5.keys())
<KeysViewHDF5 ['index', 'index_2', 'input_brand', 'input_category', 'input_composition', 'input_concat_description', 'input_department', 'input_description', 'input_gender', 'input_image', 'input_msrpUSD', 'input_name', 'input_pose', 'input_productID', 'input_season', 'input_subcategory']>

It seems there isn't fashion matching data. Where is the that information?

Hi,

Did you solve the issue?

Hi,

Did you solve the issue?

No, I haven't. But Polyvore and Amazon dataset are ok.