Livestreaming is so hot right now that it satisfies people’s desire for deeper communication through the Internet. In addition to technology and machines, the cost of making a live broadcast website is mainly network bandwidth. The clearer it is, the higher the cost of bandwidth will be. The price of the machine with 3M bandwidth on Aliyun is 3-4 times that of 1M bandwidth, while the bandwidth required for video broadcast is at least 10M or 20M, which is a very bandwidth consuming application.
But one thing I found interesting recently is that live streaming can actually be faked. Through the depth of image style conversion, you can change a rough man into a fragrant beauty, this is the power of technology ah, ha ha 😄
If it doesn’t work, check out the video here.
The original video is on the left, and the algorithmically transformed video is on the right.
The algorithm, called CycleGAN, transforms each pixel with a generative adversative network, and the upper layer is implemented with Keras/Torch/Lasagne: Tjwei /GANotebooks
So we can imagine, in fact, as traffic and bandwidth get cheaper and cheaper in the future, there will be more and more scenarios where we can communicate with people via Internet video. Until now, we have seen a lot of anonymous forgery, identity fraud and other things, because it is difficult for us to identify who is behind the text, so the Internet users in the early days of the speech mostly to vent their dissatisfaction with life.
Live video may seem more realistic, but it can also be faked, and it’s hard to spot it. Will computers be able to calculate what we want to see in the future? I think the future video is not limited to the current flat screen, there will be 3D holographic, VR, more “feel”, but all calculated, since it is calculated, it may be “fake”.
But then again, do we really care that much about “true” and “false”? You don’t know what you don’t know. It is hard to tell what is true from what is false, so there is no longer any truth from what is false, and the criteria will change.
I think this kind of technology, since it is seen by ordinary people, should have been used to make profits long ago. For example, many video websites and porn websites need a lot of video content every day, and this can reduce the cost of producing video content. It is a perfect business. Again, the need to produce drives the development of science and technology, while human desire drives the development of production.
CycleGAN
Unpaired Image-to-Image Translation using Cycle-Consistent Adversarial Networks