Skip to content

Latest commit

 

History

History
20 lines (14 loc) · 659 Bytes

HOW_TO_PACK_DATASET.md

File metadata and controls

20 lines (14 loc) · 659 Bytes

How to process the dataset

Digit datasets

Download the digits from the original sites. Due to the license, this repository does not provide any format of data.

For digits, we provide the following sample code to pack the data used in this repository, assuming that the dataset is stored under the folder datasets.

  • MNIST

    $ python tools/pack_dataset.py --root datasets/MNIST --dataset mnist --dump_path datasets/MNIST.bin

  • MNIST-M

    $ python tools/pack_dataset.py --root datasets/MNIST-M --dataset mnistm --dump_path datasets/MNIST-M.bin

Office31 dataset

VisDA dataset