Closed rajeshjoshi8792 closed 2 years ago
Hi @rajeshjoshi8792 I researched your question extensively, but most facial landmark detection systems that came up in my research involved use of RGB in generating landmarks, with depth used for anti-spoofing measures. I did though locate a research paper that discusses making use of RealSense D435 and depth-only data for facial analysis.
https://www.sciencedirect.com/science/article/pii/S1077314221000916
@MartyG-RealSense How i use depth data for anti-spoofing measurements.?
The RealSense SDK provides a C++ example for doing so with its Dlib face authentication wrapper, though the example makes use of both RGB and depth data.
RealSense Dlib wrapper https://github.com/IntelRealSense/librealsense/tree/master/wrappers/dlib
Dlib 'rs-face-dlib' anti-spoofing example program https://github.com/IntelRealSense/librealsense/tree/master/wrappers/dlib/face
@MartyG-RealSense thank you for reply. how can i use this code in python? is there python wrapper code available for anti spoofing?
The Dlib face detection library was made for C++ but supports Python. If you go to the link below and find the Help / Info section in the side-panel, expanding open the Examples: Python section shows a list of Dlib face sample programs for Python.
A RealSense team member in another case also recommended the Dlib approach, suggesting in https://github.com/IntelRealSense/librealsense/issues/4430#issuecomment-511660804 to google for the term facial-landmarks-dlib-opencv-python
@MartyG-RealSense i checked above links. but its using RGB image only. i want to use depth data.
Another suggestion about anti-spoofing with depth information is given at https://github.com/IntelRealSense/librealsense/issues/3451#issuecomment-471420572
Hi @rajeshjoshi8792 Do you require further assistance with this case, please? Thanks!
Case closed due to no further comments received.
im working on face authentication. and i can able to authenticate face using RGB image. but now i want to authenticate face using depth data. so is there any way that i can obtain facial landmark directly from depth data. i dont want to use RGB data for landmark detection.,