Repository for personal work on a MSc dissertation at Cardiff University. Program Data Science and Analytics. The purpose is to review and study the current work and publications in the field of image forensics, in particular PRNU strategies and their relation with deep learning techniques.
Dresden and Korus datasets, containing pictures from identified camera devices, and tampered pictures.
The respository is structured as follows:
classesPackage containing all wrappers used for this projectscripts/one-off scripts for dataset manipulationdocs/dissertation report done with RStudio and Markdowndatasets/Korus and Dresden dataset pictures
- For a detailed description of the experiments and results obtained please refer to our report.
To run workloads, use the following command:
python main.py --instruction run_workloads --wl_state <optional_workload_state> --wl_ID <optional_workload_ID> --experiment <optional_experiment_ID> wl_state: Optional. Specify the state of the workloads to run (default is 'awaiting'). wl_ID: Optional. Specify the ID of a specific workload to run. experiment: Optional. Specify the ID of a specific experiment.
Find Workload To find a specific workload, use the following command:
- python main.py --instruction find_workload --wl_ID <workload_ID> wl_ID: Required. Specify the ID of the workload to find.
Generate Workloads To generate new workloads, use the following command:
- python main.py --instruction generate_workloads --experiment <experiment_ID> experiment: Required. Specify the ID of the experiment for which workloads should be generated.
Delete Workloads by Experiment To delete workloads based on an experiment, use the following command:
- python main.py --instruction delete_workloads_by_experiment --experiment <experiment_ID> experiment: Required. Specify the ID of the experiment to delete workloads.
Usage Example Here's an example to generate workloads:
- python main.py --instruction generate_workloads --experiment 7 This command generates workloads for experiment 7.
- See our report.