Skip to content

Latest commit

 

History

History
96 lines (85 loc) · 5.07 KB

README.md

File metadata and controls

96 lines (85 loc) · 5.07 KB

Dataset Corrections for Common Image Manipulation Localization Datasets

English | 简体中文

1 Introduction

Image Manipulation Localization (IML) task involves detecting and locating tampered regions within images, which can be regarded as a countermeasure against methods like Photoshop or Deepfake.

However, there are many issues or resolution misalignment in the existing datasets, meaning the manipulated images and their corresponding masks do not have the same resolution. This is problematic, so this repository has addressed this issue present in various datasets.

For instance, here are some examples of problematic images in the CASIAv2 dataset:

Thus, we collected the images with this issue and revised the corresponding mask.

2 Issues & Corrected Download Links

We also point out some minor errors here, and welcome anyone to raise issues or submit pull requests to share various problems existing in the IML dataset, so that the community can work together to solve them.

2.1 CASIAv1.0 dataset

  • Issue: There is an extra image (CASIA1.0/Modified Tp/Tp/Sp_D_NRN_A_cha0011_sec0011_0542.jpg) without a mask in the CASIAv1 dataset
  • Solution: We recommend removing it during training or evaluation.

2.2 CASIAv2.0 dataset

  • Issue: There are 17 images with resolution misalignment problems.
    • File name of these images and the resolution of images & masks:
      [["Tp_D_CNN_M_N_sec00011_cha00085_11227.jpg", [256, 384, 3], [384, 256, 3]], 
      ["Tp_D_CRN_S_N_ani10191_ani10190_12437.jpg", [638, 336, 3], [336, 638, 3]],  
      ["Tp_D_CRN_S_N_nat10130_pla00049_11524.jpg", [256, 384, 3], [384, 256, 3]], 
      ["Tp_D_NND_M_B_nat20098_nat20073_01602.tif", [387, 581, 3], [382, 581, 3]], 
      ["Tp_D_NRN_M_N_nat10134_nat00095_11912.jpg", [600, 600, 3], [475, 600, 3]], 
      ["Tp_D_NRN_M_N_nat10134_nat10124_11913.jpg", [600, 600, 3], [475, 600, 3]], 
      ["Tp_S_CRN_S_N_art00059_art00059_10508.tif", [256, 384, 3], [384, 256, 3]], 
      ["Tp_S_NND_S_N_sec20064_sec20064_01654.tif", [647, 416, 3], [636, 416, 3]], 
      ["Tp_S_NNN_S_N_art20077_art20077_01883.tif", [867, 578, 3], [864, 573, 3]], 
      ["Tp_S_NNN_S_N_ind20037_ind20037_01778.tif", [578, 863, 3], [569, 862, 3]], 
      ["Tp_S_NNN_S_N_sec00012_sec00012_11230.jpg", [256, 384, 3], [384, 256, 3]], 
      ["Tp_S_NNN_S_N_sec00074_sec00074_00751.tif", [384, 256, 3], [384, 255, 3]], 
      ["Tp_S_NRD_S_N_arc20079_arc20079_01719.tif", [392, 591, 3], [383, 582, 3]], 
      ["Tp_S_NRD_S_N_pla20071_pla20071_01971.tif", [501, 760, 3], [499, 760, 3]], 
      ["Tp_S_NRN_S_B_ind10002_ind10002_20010.jpg", [600, 450, 3], [800, 600, 3]], 
      ["Tp_S_NRN_S_N_art20077_art20077_02316.tif", [863, 574, 3], [863, 572, 3]], 
      ["Tp_S_NRN_S_N_pla20080_pla20080_01980.tif", [781, 514, 3], [781, 512, 3]]]
      
  • Solution: We fixed them and released the download link for these images.
    • For only revised images, download from Google Drive.
    • For only revised images, download from Baidu Netdisk.
    • For the full revised dataset, go to this repo.

2.3 COVERAGE dataset

  • Issue: There are 9 images (27 masks) with resolution misalignment problems.
    41copy.tif
    41forged.tif
    41paste.tif
    48copy.tif
    48forged.tif
    48paste.tif
    55copy.tif
    55forged.tif
    55paste.tif
    56copy.tif
    56forged.tif
    56paste.tif
    57copy.tif
    57forged.tif
    57paste.tif
    58copy.tif
    58forged.tif
    58paste.tif
    59copy.tif
    59forged.tif
    59paste.tif
    61copy.tif
    61forged.tif
    61paste.tif
    95copy.tif
    95forged.tif
    95paste.tif
    
  • Solution: We fixed them and released the download link for these images.

2.4 IMD2020 dataset

  • Issue: There is a single image(IMD2020/z14/00030_fake.jpg) with a resolution misalignment problem.
  • Solution: We simply upload the revised mask here, you can just download it directly:

3 Supports and Sharing

  • If you come across any other issues in datasets within the IML field, feel free to point out them in the Discussions and let the open-source community work together to resolve them.
  • If you find our work valuable, please consider giving us a star⭐️ and sharing it with others. Your support helps us gain more recognition and encourages further collaboration within the community.

Flag Counter