# Detecting Photoshopped Faces by Scripting Photoshop

**Authors:** Sheng-Yu Wang, Oliver Wang, Andrew Owens, Richard Zhang, Alexei A. Efros  
**arXiv:** [1906.05856](https://arxiv.org/abs/1906.05856)  
**Date:** 2019-06-13  
**Summary page:** https://www.summarizepaper.com/en/paper/1906.05856v2/

## Key points

- Researchers address malicious photo manipulations using Adobe Photoshop
- Introduce novel method for identifying image warping on human faces
- Model trained on synthetic images generated through automated scripting in Photoshop
- Model surpasses human performance in recognizing manipulated images and predicting alterations
- Enables reversal of edits to reconstruct original unaltered image in some cases
- Effective in detecting digitally altered images and artist-created manipulations
- Holds promise for enhancing digital forensics capabilities and combating deceptive practices

## Summary

In their paper titled "Detecting Photoshopped Faces by Scripting Photoshop," authors Sheng-Yu Wang, Oliver Wang, Andrew Owens, Richard Zhang, and Alexei A. Efros address the prevalent issue of malicious photo manipulations carried out using popular image editing tools like Adobe Photoshop. The researchers introduce a novel method for identifying a specific type of manipulation commonly seen in Photoshop edits – image warping on human faces. Their approach involves training a model exclusively on synthetic images generated through automated scripting within Photoshop itself. Remarkably, the developed model surpasses human performance in recognizing manipulated images and accurately predicting the precise locations of alterations. It even enables the reversal of edits to reconstruct the original unaltered image in certain instances. The team's findings demonstrate the efficacy of their system not only in detecting digitally altered images but also in successfully identifying and addressing real-world artist-created manipulations. This innovative technique holds significant promise for enhancing digital forensics capabilities and combating deceptive practices in visual media. The research sheds light on the potential applications of automated tools in safeguarding the integrity and authenticity of digital imagery amidst an increasingly sophisticated landscape of photo manipulation techniques.