larger dataset

[redwangwangwang]

Thank you for your great work! Have you ever done experiments on larger dataset (for example, 10000 for each class as a training set) or in quickdraw_origin? If not, can you provide the method of making the dataset? Thank you again!

[w-haiS]

Thank you for your great work! Have you ever done experiments on larger dataset (for example, 10000 for each class as a training set) or in quickdraw_origin? If not, can you provide the method of making the dataset? Thank you again!

I had the same problem, did you solve the problem of how to make the data set?

[redwangwangwang]

Here is the code I adapted and used: https://github.com/redwangwangwang/pg2023. You need to download the file in QuickDraw-50M npz format first, and then convert it to hdf5 format. Then use the Dataset file to read. Note that you may need to simply rewrite this since I added an extra cls token. If it works for you, please give a star.