Researchers can use the data to train automatic fake recognition algorithms.
Google has released a database with thousands of deephakes to support developers creating fake detection tools. To do this, the company shot hundreds of videos with actors and made deephakes using publicly available methods. This was reported by Google employees on the company’s blog.
Researchers can use the Google dataset to train automatic deepfake recognition systems. Thanks to the data set, developers will be able to increase the accuracy and efficiency of tools as much as possible. Google promised to add new videos to the database to cope with the development of algorithms for generating deepfakes.
As the region is developing rapidly, we will supplement the dataset with the development of technology of deepfakes, and we will continue to work with partners in this field. We strongly believe in supporting a community of researchers fighting the potential dangers of synthetic media.
from google blog
Google is not the only technology company to oppose deepfakes. Facebook and Microsoft are also participating in a joint initiative to create tools for finding deepfeiks. Companies are developing open source programs that will help governments, companies, and the media identify fraudulent videos.