search

cheind / dest

:panda_face: One Millisecond Deformable Shape Tracking Library (DEST)
BSD 3-Clause "New" or "Revised" License
286 stars 109 forks source link

Using in iOS app #11

Closed tveritinova closed 6 years ago

tveritinova commented 7 years ago

Hello!

Can I use this library in iOS application? If yes, how?

cheind commented 7 years ago

Hi! I'm not really familiar with iOS, but I'd say this should be possible.

One way to go about it is to use iOS built-in face detection and then run DEST on rectangles provided by iOS to regress the landmarks. To make this work you will need to train DEST on typical rectangles provided by the face detection algorithm. This could be accomplished by running the face detection of iOS (maybe in simulation on a Mac) on all training and test images to provide rough estimates of faces. Then train DEST on these images / rectangles. The resulting regressor is stored as file that can be uploaded to the mobile phone (and is binary portable). Finally wrap the core of DEST using your language of choice on iOS. I'm pretty sure that Swift allows you to call C++ code from within Swift.

Alternatively, you can use OpenCV on iOS and use the pretrained model files.