A PyTorch implementation of the DeLR (Dual-encoder Landmark Regression) architecture for cephalometric landmark detection, evaluated on three public datasets:
Aariz Cephalograms — 1000 images, 29 annotated landmarks (700 / 150 / 150 train/valid/test).
CephAdoAdu Dataset — 700 images, 10 landmarks, mixed adolescent + adult cohort (400 train / 300 test in the official splits; we held out 10 % of train as validation).
ISBI 2015 Cephalometric… See the full description on the dataset page:
https://huggingface.co/datasets/emad2001/DeLR-Cephalometric-ConvNeXtV2.
