The Photoshop Paradox: Using Fake Images to Detect Real Manipulations
Shahnab Ahmed
Product Development | GAI Enthusiast | Media Ace Rising Star 2019 | e4m40under40 Award Winner 2019 | National Institute of Technology, Silchar
"The Photoshop Paradox: Using Fake Images to Detect Real Manipulations."
Yes, you read that right. The researchers used fake images generated by scripting Photoshop to train a model capable of detecting image warping applied to human faces. Clearly, the Photoshop gods have a sense of humor.
Who knew that the key to detecting image manipulation was hiding in plain sight all this time? The dynamic duo of Adobe and UC Berkeley decided to put Photoshop to the test, scripting fake images to train a model capable of detecting image warping techniques used to manipulate human faces.
The algorithm can sniff out manipulations, predict their location, and even undo them, leaving you with a boring, unedited photo. Bummer.
Read more about the project here- https://peterwang512.github.io/FALdetector/
Personally, have implemented the repo in Google Colab- can try the following link https://colab.research.google.com/drive/1Y7k6DvDBYhkyRpWFU-_Rdvo832EZb9wc?usp=sharing
Streamlit Execution for the codes:
#ProjectAboutFace : Adobe MAX (Sneak Peek) | Adobe Creative Cloud