# faceswap

**Repository Path**: minskyli/faceswap

## Basic Information

- **Project Name**: faceswap
- **Description**: No description available
- **Primary Language**: Unknown
- **License**: Not specified
- **Default Branch**: master
- **Homepage**: None
- **GVP Project**: No

## Statistics

- **Stars**: 0
- **Forks**: 1
- **Created**: 2018-10-24
- **Last Updated**: 2020-12-19

## Categories & Tags

**Categories**: Uncategorized

**Tags**: None

## README

---

# deepfakes_faceswap

Faceswap is a tool that utilizes deep learning to recognize and swap faces in pictures and videos.

## introduction
we do four experiments in this project, you can find experiments-record word file, train data, target data(videos:frame), experiment videos in the Enterprise cloud disk of our team.
you can download the train data in exp_1/exp_2 in https://anonfile.com/p7w3m0d5be/face-swap.zip.

## overview
The project has multiple entry points. You will have to: extract, train, convert.

### extract
at first, you can use  `ffmpeg -i ../trump.mp4 -vf fps=30  -qscale:v 2 "trump1%04d.jpg` to convert your video in image.
and then, you can  `python faceswap.py extract -h`to see what args you can use and change, now you can extract the faces in train data
`python faceswap.py extract`,and don't forget to change which GPU you want to use, you can change that in `faceswap.py`.

### train
now, you can train your data `nohup python faceswap.py train -A ./ae/ac/fallon -B ./ae/ac/trump -m ./models/ae/ac -s 100 -t Original -bs 64 -it 400000 -g 1 -pl > logs.txt 2>&1 &`

### convert
finally, you can convert the face A to face B `nohup python faceswap.py convert -i ./target_data/cage_target/ -o ./transfer_data/cage/ -m ./models/exp_4/models/ae/ab/ -c Masked -t Original -b=9 -D mtcnn -M facehullandrect -g=1 > logs_gan_ab.txt 2>&1 &`
and ues `ffmpeg  -f image2 -i cage%04d.jpg -vcodec libx264 -crf 15 -pix_fmt yuv420p ./cage_ae_ac_b9.mp4`to convert images to a video.

## tip
you can see what we did before in our Enterprise cloud disk, the file name is **video faceswap experiments**
this project has a relevant paper named **Fast Face-swap Using Convolutional Neural Networks**