Week 6 — Copy Move Forgery Detection
Hello, in this week we completed the porting of the MATLAB algorithm to Python and ran some test cases in the program to get some outputs.
Dataset
Since our project was established upon an existing paper, we have different datasets that we can work with, the dataset that we used for our small example can be downloaded here. It has 220 images in it and a ground truth file that describes them as 110 original and 110 forged images.
Our first task was to extract SIFT features on images to compare them. An example can be found in the images below.
We tried with many threshold values to minimize False Positives and Maximize True Positives which will be used as our evaluation metric through the project, we achieved a relative success with both values individually but not a great value that achieves a satisfactory result for both of them at the same time.
Our next goal is to change our matching and post processing algorithms in the dataset to achieve a better success.