Crypto News

Image Manipulation Taking a Different Shape with Deepfake Emotions

Deepfake is quite common these days for depicting people saying or doing that they never did. It is a photo, video, or audio recording that seems real but has been manipulated using artificial intelligence or AI. The underlying technology can manipulate facial expressions, replace faces, synthesize speech and faces. And so recently Microsoft has also come with a deepfakes tool that can identify image manipulation and facial manipulation.


Any ideas on how the deepfake works with regard to image manipulation?

Most of the deepfake videos are of swapped faces or manipulated facial expressions. Since they rely on AI neural networks that can recognize the patterns of data, AI needs to be fed with many images to train it and to identify and reconstruct the usual faces.

Usually, deepfakes are used in movies to give a creative effect using image manipulation. But it also has great risks too, as they are also used for exploitation. Few studies showed that deepfake content online is pornographic and can misappropriate victimize women.

There is also a concern about the use of deepfakes for disinformation. It can also be used to influence elections or as a weapon of psychological warfare and also lead to disregard of legitimate evidence of wrongdoing to undermine public trust.


Image manipulation and deepfakes

Image manipulation is a technique that allows anyone to alter images or even videos to swap identities. The most popular techniques are Deepfakes and Face2Face which have gained a lot of popularity these days. The ethical factors around the utilization of DeepFakes and Face2Face come into the picture too.

The great news is that a recent journal paper at Hunan and Nanjing University in China has proposed a new method for identifying and detecting DeepFake manipulation through an Adaptive Manipulation Traces Extraction Network (AMTEN).

AMTEN is a method that works as a pre-processing module for combating the side effects of image manipulation. The new technique uses the difference between the input image and the output feature map to extract these manipulation traces. The AMTEN method seems to work well to detect manipulation traces that tend to tamper with detection tasks.

Facial Image Manipulation techniques can be divided into three categories: identity manipulation, expression manipulation, and attribute transfer. AMTENnet considers multiple classifications of facial image manipulation techniques. AMTENnet has also resulted in increased accuracy for detecting the deepfakes in image manipulation.

Share This Article

Do the sharing thingy

About Author

More info about author

Source link

Related Articles

Leave a Reply

Your email address will not be published.

Back to top button