the code for our WACV2025 paper "Cascaded Dual Vision Transformer for Accurate Facial Landmark Detection"

train

download the pre-cropped WFLW images from https://github.com/starhiking/HeatmapInHeatmap?tab=readme-ov-file, unzip the WFLW.zip under root folder and run the following commands:

    torchrun --nproc_per_node=2 TrainHeatmapStageFP16.py

Name		Name	Last commit message	Last commit date
Latest commit History 2 Commits
loss		loss
AUGHIH.py		AUGHIH.py
Attention.py		Attention.py
Config.py		Config.py
Data300WAngleWeights.py		Data300WAngleWeights.py
DataMutliResHM.py		DataMutliResHM.py
Dataset.py		Dataset.py
Dataset2.py		Dataset2.py
Dataset300W.py		Dataset300W.py
DatasetAAM.py		DatasetAAM.py
DatasetAll.py		DatasetAll.py
DatasetCOFW.py		DatasetCOFW.py
DatasetLHM.py		DatasetLHM.py
DatasetMicroWFLW.py		DatasetMicroWFLW.py
DatasetTagent.py		DatasetTagent.py
DrawHeatmap.py		DrawHeatmap.py
EMA.py		EMA.py
Evaluation.py		Evaluation.py
EvaluationAll.py		EvaluationAll.py
EvaluationVideo.py		EvaluationVideo.py
Heatmap.py		Heatmap.py
Hourglass.py		Hourglass.py
ImageAugmentation.py		ImageAugmentation.py
Net.py		Net.py
README.md		README.md
RandomFlip.py		RandomFlip.py
Regression.py		Regression.py
TrainHeatmapStageFP16.py		TrainHeatmapStageFP16.py
UNet2.py		UNet2.py
UNet3.py		UNet3.py
UNetHM.py		UNetHM.py
Vit.py		Vit.py
ccnet.py		ccnet.py
coord_conv.py		coord_conv.py
helpers.py		helpers.py
loss_function.py		loss_function.py
main.py		main.py