PAL-Net Enables Accurate 3D Facial Landmark Localization

PAL‑Net, a point‑wise CNN with patch‑attention, localizes 3‑D facial landmarks with a mean error of 3.686 mm on 214 adult scans, and achieves 0.41 mm error on the large FaceScape dataset. getnews.me/pal-net-enables-accurate... #palnet #3dfacial #landmarking