datadings.sets.Places365_write module

usage: Places365_write.py [-h] [-o PATH] [-y] [-s] [-l] [-c] INDIR

Create Places365 data set files. You can choose between any combination of

  • high or low resolution images and

  • standard or challenge (extended) training data set

This tool will look for the following files in the input directory depending on the chosen version and download them if necessary:

  • standard (large)

    • train_large_places365standard.tar (105 GB)

    • val_large.tar (2.1 GB)

    • test_large.tar (19 GB)

  • standard (small)

    • train_256_places365standard.tar (24 GB)

    • val_256.tar (501 MB)

    • test_256.tar (4.4 GB)

  • challenge (large)

    • train_large_places365challenge.tar (476 GB)

    • val_large.tar (2.1 GB)

    • test_large.tar (19 GB)

  • challenge (small)

    • train_256_places365challenge.tar (108 GB)

    • val_256.tar (501 MB)

    • test_256.tar (4.4 GB)

Important

For performance reasons shuffling is not available. It is recommended to use the datadings-shuffle command to create a shuffled copy.

Important

Samples have the following keys:

  • "key"

  • "image"

  • "label"

Positional arguments

INDIR Directory that contains dataset source files.

options:
-h, --help

show this help message and exit

-o PATH, --outdir PATH

Output directory. Defaults to indir.

-y, --no-confirm

Don’t require user interaction.

-s, --skip-verification

Skip verification of source files.

-l, --low-res

Download the resized and cropped images (256x256). Default are images with minimum dimension of 512 and preserved aspect ratio.

-c, --challenge

Download the extended challenge training dataset. Validation and testing are the same.

datadings.sets.Places365_write.file_spec(path, md5, total=None)[source]
datadings.sets.Places365_write.get_files(challenge, low_res)[source]
datadings.sets.Places365_write.main()[source]
datadings.sets.Places365_write.write_sets(indir, outdir, args)[source]