DAD-3DHeads: A Large-scale Dense, Accurate and Diverse Dataset for 3D Head Alignment from a Single Image
Please refer to original readme of the repo to setup the DAD-3DHeads. I have modified this read for my work
I intended to use DADNet for dense 2D landmark detection. It performs well on DAD-3DHeads dataset but performance deteriorates when face is only a part of the image. Check Demo section. I used mediapipe facial detector and then passed the detected face to DADNet. The predicted 2D landmarks are then projected back on the original image. See demo section for details
- Step1: Set up DADNet using the intructions here
- Step2: pip install mediapipe (pip install mediapipe-silicon for mac silicon)
Check this notebook
