Thank you so much for your repo!

[denisenepraunig]

Hello Taehee Lee!

I stumbled upon your repository and it sparked my interest in using Keras. I am an iOS developer and have little Machine Learning experience. After some effort and using Google Colab I was able to convert your landmarks h5 model to a CoreML model (with coremltools) and it predicted the correct landmarks inside an iOS App of a cat face 🎉!

Thank you so much for your effort and putting everything into this repo.

I want to build a snapchat like app where the cat would also be wearing glasses. Let's see how far I will come 😅.

Have a great day!!

Best regards,
Denise

[flankechen]

@denisenepraunig grad to take a glance of your work, nice! how far you have came? are you trying this on native code or HTML5 web things

[denisenepraunig]

I am was building this with native for iOS with Swift. Currently I was able to highlight the landmarks positions inside the app. Still need to do some math then for the angle of the glasses and put them over the cat face :D I will put this into a repo when the code is cleaned up and write some short documentation about it!

[flankechen]

@denisenepraunig cool, about "putting stickers on the face", I used to achieve like, graphic designer make some cute sticker on a standard cat face, output a onlyt sticker png with transparence.
when input image landmarks are detected, trying to find a transform from starndard cat face landmark to current input cat face landmark. and translation, rotation and scale can be extracted from that transform.