Very recent fire deepfake, there is a very recent fire in a dance now is a good application for GAN, and the effects are very good
You can talk about what the principle is, both the principle of some distinction, and then talk about the difference between where you are, if you can, you can do it yourself a demo
Recently we have seen a lot of reports about Yang Mi face transplant Athena, which use the technology called deepfake
The original core deepfake is this:
Then he is a model of self-decoding,
The training process is distorted A A becomes untwisted, i.e. a distorted input data A, and then through the encoder and decoder processes two,
extracting feature encoder, decoder is mapped on the image feature
Then the technical steps divided into three parts:
1 get the picture, such as Yang Mi's face: This should be digging through dlib Recognition Library inside face recognition (he can get coordinates 36 key points inside a picture of a human face, so you can put a face to pull out )
Next we look at in terms of what is dance ai
This dance ai is doing?
dance ai function is given two videos are A video (a video Yang Ying) and B video (for a standard human prime yoga video), we want to get a standard Yang Ying doing yoga video
Very recent fire deepfake, there is a very recent fire in a dance now is a good application for GAN, and the effects are very good
You can talk about what the principle is, both the principle of some distinction, and then talk about the difference between where you are, if you can, you can do it yourself a demo
Recently we have seen a lot of reports about Yang Mi face transplant Athena, which use the technology called deepfake
The original core deepfake is this:
Then he is a model of self-decoding,
The training process is distorted A A becomes untwisted, i.e. a distorted input data A, and then through the encoder and decoder processes two,
extracting feature encoder, decoder is mapped on the image feature
Then the technical steps divided into three parts:
1 get the picture, such as Yang Mi's face: This should be digging through dlib Recognition Library inside face recognition (he can get coordinates 36 key points inside a picture of a human face, so you can put a face to pull out )
Next we look at in terms of what is dance ai
This dance ai is doing?
dance ai function is given two videos are A video (a video Yang Ying) and B video (for a standard human prime yoga video), we want to get a standard Yang Ying doing yoga video