Computer-Aided Annotation for Video Tampering Dataset of Forensic Research

TL;DR

This paper introduces a computer-aided annotation method for efficiently labeling forged video frames, combining manual input with predictive algorithms to improve accuracy and speed in video tampering datasets.

Contribution

The proposed method enhances annotation efficiency and accuracy for forged videos by integrating manual marking with linear prediction algorithms, addressing limitations of manual and automated annotation techniques.

Findings

Reduces annotation time for forged videos.

Improves accuracy of forged area localization.

Facilitates large-scale video forensic research.

Abstract

The annotation of video tampering dataset is a boring task that takes a lot of manpower and financial resources. At present, there is no published literature which is capable to improve the annotation efficiency of forged videos. We presented a computer-aided annotation method for video tampering dataset in this paper. This annotation method can be utilized to label the frames of forged video sequences. By means of comparing the original video frames with the forged video frames, we can locate the position and the trajectory of the forged areas of the forged video frames. Then, we select several key points on the temporal domain according to the trajectory of the forged areas, and mark the forged area of the forged frames in the key point with a mouse. Finally, we use the linear prediction algorithm based on the coordinates of the key positions in the temporal domain to generate the annotation information of forged areas in other video frames which without manually labeled. If the bounding box generated by the computer-aided algorithm deviates from the actual location of the forged area, we can use the mouse to change the position of the bounding box during the preview period. This method combines the manual annotation with computer-aided annotation. It solves the problems of the inaccuracy of annotation by computer-aided as well as the low efficiency of annotation manually, and meet the needs of annotation for an enormous amount of forged videos in the research of video passive forensics.