Image Manipulation Datasets

This python package contains torch dataset classes for common image manipulation datasets.

Currently, the supported datasets are:

• CASIA
  • CASIA 2.0
• Defacto
  • Copy/Move
  • Splicing
  • Inpainting
• Coverage
• IMD2020

Install

pip install git+https://github.com/cainspencerm/[email protected]

Examples

CASIA 2.0

Ensure that the ground truth directory is in data_dir and named 'CASIA 2 Groundtruth'.

import image_manip.datasets as imds

# Create dataset object for dataloader.
dataset = imds.Casia2(data_dir='data/CASIA2.0')  # optional split=['train', 'val', 'test', 'benchmark', 'full']

Defacto Copy/Move

import image_manip.datasets as imds

# Create dataset object for dataloader.
dataset = imds.CopyMove(data_dir='data/copy-move')  # optional split=['train', 'val', 'test', 'benchmark', 'full']

Sample Quality

Datasets are not always perfect. Of the available datasets, COVERAGE, CASIA 2, and Defacto Splicing had images and masks that didn't match in size, though they have been verified as pairs. For this reason, the dataset classes resize the masks to the size of the original image, with the hopes that the masks line up correctly with the image. This is unverified as it would require manually verifying each of the over 110,000 image and mask pairs.