Facial landmarks become a necessary task for some applications just like morph or swap faces and you can't get a good result without getting the exact face landmarks to keep sure the accuracy is always maintained!. Also to keep the performance highly speed we used trackers which we support more than 5 type of them described below instead of detect faces multiple times which will take down a massive of performance.
The project depends on OpenCV Library you can check it here opencv
- Here we tried this on a photo of the CS50 course instructors.
🔥 Youtube Video
We just put the lib for only armeabi-v7a mobiles's CPU you can make X86 or other ABI as well
- Download the project into your PC!
- Open it using android studio and try to build it.
- Put your preffered tracker you can choose one of those
- MedianFlow
- BOOSTING
- TLD
- KCF
- CSRT
- MIL
- MOSSE
- GOTURN (Based on Convolutional Neural Network)
More information can be found here for each tracker Here
- Run your code!
- Cropping the exact face from the image
feel free to contribute if you're familiar with Image processing
MIT