Automatic Seamless Face Replacement in Videos

-Yiren Lu (luyiren [at] seas [dot] upenn [dot] edu) -Dongni Wang (wdongni [at] seas [dot] upenn [dot] edu)

Automatic Seamless Face Replacement (without deep learning).

Project Description

Python third party libs required:

To run face replacement:

$ wget http://dlib.net/files/shape_predictor_68_face_landmarks.dat.bz2
$ tar xvfj shape_predictor_68_face_landmarks.dat.bz2

$ Python face_detect_wrapper.py shape_predictor_68_face_landmarks.dat [video_name_no_suffix]

e.g.:

$ Python face_detect_wrapper.py shape_predictor_68_face_landmarks.dat clips/clip1

Example outputs in Proj4_Test/ and clips/

Face replacement: see demo.m
- load face detection results output by 1.
- run replace_all_faces([src video path], [replacement video path], [src video detection results], [source video detection results], [destination video detection results], [source face index], [resize x])
- save video to .avi file

Example output videos in output_videos/

The contents of this repository are licensed under the MIT License.

Name		Name	Last commit message	Last commit date
Latest commit History 20 Commits
Proj4_Test/easy		Proj4_Test/easy
blending		blending
carving		carving
clips		clips
morphing		morphing
utils		utils
.gitignore		.gitignore
README.md		README.md
crop_face.m		crop_face.m
demo.m		demo.m
face_detect_wrapper.py		face_detect_wrapper.py
face_replacement.m		face_replacement.m
face_replacement_beta.m		face_replacement_beta.m
face_shape_feat.m		face_shape_feat.m
face_tracking.m		face_tracking.m
face_traj_smoother.m		face_traj_smoother.m
generate_test_videos.m		generate_test_videos.m
match_faces_ransac.m		match_faces_ransac.m
project_description.pdf		project_description.pdf
replace_all_faces.m		replace_all_faces.m
video_screenshot.jpg		video_screenshot.jpg