156 points by sarah-codes 5 months ago flag hide 13 comments
johnsmith 5 months ago next
Great work! I've been looking for a real-time face tracking solution and this looks perfect.
user2 5 months ago next
Nice! I'm curious how well it performs on mobile devices?
johnsmith 5 months ago next
@user2 Haven't tried it on mobile devices yet, but I'll definitely test it out and let you know the results.
ml_fan 5 months ago prev next
Did you use TensorFlow Lite for this project?
johnsmith 5 months ago next
@ml_fan No, I used the full version of TensorFlow for this demo. But I'm planning to test it with TensorFlow Lite soon.
opencv_expert 5 months ago prev next
Why not use OpenCV for face tracking? It's a lot faster and has a lot less dependencies than TensorFlow.
johnsmith 5 months ago next
@opencv_expert That's a good question. I started this project a while ago and I was already familiar with TensorFlow. But I'll definitely give OpenCV a try next time.
coder123 5 months ago prev next
This is great, I'm going to use it for my real-time face recognition project!
johnsmith 5 months ago next
@coder123 Awesome, I'd love to see what you build with it!
alexn 5 months ago prev next
This is a really impressive demo. I'm curious how you handled the face tracking in 3D?
johnsmith 5 months ago next
@alexn Thank you! I followed the same approach as the original TensorFlow face tracking tutorial. You can check it out here: <https://github.com/tensorflow/models/tree/master/research/object_detection>
anonymous 5 months ago prev next
Did you compare the performance of this solution to other face tracking algorithms?
johnsmith 5 months ago next
@anonymous Yes, I did. This solution was the most accurate and real-time, compared to other algorithms like dlib, OpenCV and Haar cascade.