home   |  research   |  members   |  projects   |  publications   |  conferences of interest   |  downloads   |  contact & impressum   |  privacy information
 
 

PLUS Temporal Image Forensics Dataset (PLUSTIFDS) - Download Page

This is "The Multimedia Signal Processing and Security Lab", short WaveLab, website. We are a research group at the Artificial Intelligence and Human Interfaces (AIHI) Department of the University of Salzburg led by Andreas Uhl. Our research is focused on Visual Data Processing and associated security questions. Most of our work is currently concentrated on Biometrics, Media Forensics and Media Security, Medical Image and Video Analysis, and application oriented fundamental research in digital humanities, individualised aquaculture and sustainable wood industry.
PLUS Temporal Image Forensics Dataset

PLUS Temporal Image Forensics Dataset

The PLUS Temporal Image Forensics Dataset (PLUSTIFDS) is a temporal image forensics dataset where content bias is limited. Usually, images taken in close temporal proximity (i.e., belonging to the same age class) share common scene properties (aka content bias). Content bias can be exploited by data-driven models instead of comprehensible evidence (i.e., age traces). With the PLUSTIFDS, the idea is to extract the artifacts and noise introduced by the image acquisition pipeline (which contains the age traces) from the acquired calibration images (i.e., Dark Field Images (DFI) and Bright Field Images (BFI)). The extracted artifacts and noise can then be embedded into synthetic (rendered) images. Synthetic images are completely free of any artifacts and noise introduced by the image acquisition pipeline. When the artifacts and noise are extracted at different points in time (age classes) they can be embedded into the same set of synthetic images. Thus, truly identical images (in terms of image content) are available per age class only with different artifacts and noise embedded.


Figure: Overview of embedding the estimated artifacts and noise (represented by Ψ and Ξ) in a synthetic image. For each age class (defined by the available calibration image acquisition sessions), the same set of synthetic images is used, but with a different artifacts and noise embedded.

This dataset could help to, (i) develop deep learning based age approximation methods (ii) facilitate the discovery of new (unknown) age traces, (iii) assess the impact of content bias on existing age approximation methods and (iv) develop and verify new eXplainable Artificial Intelligence methods. DFIs are images in which the camera’s shutter is closed so that the incident light is set to zero (I = 0), i.e., DFIs are not affected by content bias. BFIs are captured by illuminating the sensor with a uniform field. To achieve this, a spotlight with a mounted softbox is used. The softbox includes an inner and outer diffuser. A total of three different types of outer diffusers are used, i.e., a diffuse acrylic glass and two different types of white fabric. To capture BFIs, the camera is pointed directly at the softbox. The images are captured in a basement room with the window completely covered (i.e., controlled light and temperature conditions).


   
Figure: Capturing setup for recording BFIs.

DFIs were captured with different combinations of ISO settings and exposure times and BFIs with different combinations of f-number and focal length. Currently, two sessions (November 2023 and June 2024) have been recorded by 8 different imagers.

Table: Overview of the available Imagers used to create the dataset.
ID Camera Model Resolution (WxH) Sensor Raw Images
PLUS-canon01 (Pc01) Canon PS A720IS 2592x1944 CCD -
PLUS-canon02 (Pc02) Canon EOS 70D 5472x3648 CMOS CR2
PLUS-fujifilm01 (Pf01) Fujifilm X100V 6240x4160 CMOS RAF
PLUS-konica01 (Pk01) KonicaMinolta Dimage Z5 2560x1920 CCD -
PLUS-nikon01 (Pn01) Nikon E7600 3072x2304 CCD -
PLUS-pentax01 (Pp01) Pentax K5 4928x3264 CMOS DNG
PLUS-pentax02 (Pp02) Pentax K5II 4928x3264 CMOS DNG
PLUS-sony01 (Ps01) Sony DSC-P200 3072x2304 CCD -


Further details can be found in:

R. Joechl and A. Uhl, "PLUS Temporal Image Forensics Dataset," 2024 IEEE International Workshop on Information Forensics and Security (WIFS), Rome, Italy, 2024, pp. 1-6, doi: 10.1109/WIFS61860.2024.10810689.

An open source repository containing the code for extracting the artifacts and noise from the calibration images as well as a dataloader that embeds the extracted signals into synthetic images can be found in: PLUS Temporal Image Forensics Dataset (PLUSTIFDS)

Filename and Directory Structure

The following directory structure applies:
  • [image type]: contains all acquiered DFIs or BFIs.
    • [imager ID]: cotains all acquiered DFIs or BFIs per imager ID (e.g., PLUS-canon02).
      • [file format]: The DFIs or BFIs are further divded into the available file formats (e.g., JPG, DNG, etc.).

DFI filenames are encoded according to the following structure:
DFI_[imager ID]_[session ID]_iso[iso setting]_expt[exposure time]_[3-digit consecutive number].JPG

BFI filenames are encoded according to the following structure:
BFI_[imager ID]_[session ID]_[diffuser ID]_[intensity setting]_f[f-number]_fl[focal length]_[3-digit consecutive number].JPG

diffuser IDs: 00 acrylic glass, 01 fabric type I and 02 fabric type II.
intensity setting: represents the light intensity setting of the spotlight (i.e., 04 and 06).


A csv file is provided for DFIs and BFIs (in the respective directory), which contains all relevant metadata for each acquired image. In addition, for each imager exists a Sqlite database in the directory './extracted_signals', which contains all extracted signals of the corresponding imager.

Obtaining the Dataset

To obtain the PLUSTIFDS you have to agree to our license agreement:
coming soon...

Please download, fill in and sign the license agreement and send it to R. Joechl. After checking the license agreement you will be provided with a download link.