WQuatNet

WQuatNet: Wide range quaternion-based head pose estimation here

• Fig. shows the difference between the narrow and full ranges.

Results visualization

• Fig. Snapshots of 8 different views from one sequence in the AGORA Panoptic val-set dataset.

Our results

• Table 1: MAE values for the CMU dataset obtained using different methods for Narrow-range angles: −90◦ < yaw < 90◦.

Method	Retrain?	Yaw	Pitch	Roll	MAE
DirectMHP	Yes	5.86	8.25	7.25	7.12
DirectMHP	No	5.75	8.01	6.96	6.91
6DRepNet	Yes	5.20	7.22	6.00	6.14
6DoF-HPE (ours)	Yes	5.13	6.99	5.77	5.96
WQuatNet (ours)	Yes	4.67	6.77	5.42	5.62

• Table 1: MAE values for the CMU dataset obtained using different methods for Full-range angles: −180◦ < yaw < 180◦.

Method	Retrain?	Yaw	Pitch	Roll	MAE
Viet et al.	No	9.55	11.29	8.32	9.72
DirectMHP	Yes	7.38	8.56	7.47	7.80
DirectMHP	No	7.32	8.54	7.35	7.74
WHENet	No	8.51	7.67	6.78	7.65
Cobo et al.	No	–	–	–	7.45
6DRepNet	Yes	5.89	7.76	6.39	6.68
6DoF-HPE (ours)	Yes	5.83	7.63	6.35	6.60
WQuatNet (ours)	Yes	6.04	7.72	6.34	6.70

• Table 2: MAE values for the AGORA dataset obtained using different methods for Narrow-range angles: −90◦ < yaw < 90◦.

Method	Retrain?	RR	Yaw	Pitch	Roll	MAE
DirectMHP	✅	E	12.10	12.86	11.43	12.13
6DoF-HPE (ours)	✅	6D	13.92	10.66	10.24	11.61
6DRepNet	✅	6D	13.03	10.51	10.25	11.26
DirectMHP	❌	E	11.19	9.89	10.46	10.51
WQuatNet (ours)	✅	Q	10.66	9.99	9.60	10.08

• Table 2: MAE values for the AGORA dataset obtained using different methods for Full-range angles: −180◦ < yaw < 180◦.

Method	Retrain?	RR	Yaw	Pitch	Roll	MAE
DirectMHP	✅	E	13.39	15.68	12.08	13.72
6DoF-HPE (ours)	✅	6D	15.07	11.74	11.04	12.62
6DRepNet	✅	6D	14.67	11.57	10.92	12.39
DirectMHP	❌	E	12.28	12.42	11.02	11.91
WQuatNet (ours)	✅	Q	12.84	11.58	10.80	11.74

• Table 3: MAE values for the BIWI and AFLW2000 datasets using different methods.

Method	R	RR	AFLW2000 Yaw	Pitch	Roll	MAE	BIWI Yaw	Pitch	Roll	MAE
HopeNet	❌	E	6.47	6.56	5.44	6.16	4.81	6.60	3.27	4.89
FSA-Net	❌	E	4.50	6.08	4.64	5.07	4.27	4.96	2.76	4.00
WHENet	❌	E	4.44	5.75	4.31	4.83	3.60	4.10	2.73	3.48
TokenHPE	❌	RM	4.36	5.54	4.08	4.66	3.95	4.51	2.71	3.72
QuatNet	❌	Q	3.97	5.62	3.92	4.50	4.01	5.49	2.94	4.15
MFDNet	❌	RM	4.30	5.16	3.69	4.38	3.40	4.68	2.77	3.62
Srnet	❌	Q	3.75	5.10	3.46	4.10	3.81	4.36	2.77	3.65
DirectMHP	❌	E	2.99	5.35	3.77	4.04	3.57	5.47	4.02	4.35
Li et al. 2022	❌	RM	3.36	5.05	3.56	3.99	3.59	3.94	2.68	3.40
6DRepNet	❌	6D	3.63	4.91	3.37	3.97	3.24	4.48	2.68	3.47
FSA-Net (retrained)	✅	E	5.41	6.82	5.42	5.88	4.74	5.32	3.26	4.44
6DRepNet (retrained)	✅	6D	3.50	4.81	3.47	3.93	3.79	4.53	2.89	3.74
6DoF-HPE (ours)	✅	6D	3.56	4.74	3.35	3.88	3.91	4.43	2.69	3.68
WQuatNet (ours)	✅	Q	3.39	4.62	3.23	3.75	3.91	4.20	2.64	3.58

• Table 4: MAE values for the BIWI dataset (split in a 70:30 ratio) using different methods.

Method	RR	Yaw	Pitch	Roll	MAE
FSA-Net	E	2.89	4.29	3.60	3.59
HopeNet	E	3.29	3.39	3.00	3.23
MFDNet	RM	2.99	3.68	2.99	3.22
SRNet	Q	3.01	2.78	2.86	2.88
6DRepNet	6D	2.69	2.92	2.36	2.66
TokenHPE	RM	3.01	2.28	2.01	2.43
WQuatNet (ours)	Q	2.39	2.37	2.08	2.28

Datasets

• CMU Panoptic from here for the full range angles.

• 300W-LP, and AFLW2000 from here for the narrow range angles.

• BIWI from here for the narrow range angles.

For Training:

If you only need to change the pre-trained RepVGG model 'RepVGG-D2se-200epochs-train.pth' please see here and save it in the root directory.

python3 train.py

After training is done. Next step.

For Deploy models:

For reparameterization, the trained models into inference models use the convert script.

python3 convert.py input-model.tar output-model.pth

After converting the training model into an inference model. Then, you can test your model.

For Testing:

python3 test.py

If you are too lazy to do the above steps, you can download the pre-trained RepVGG model to run your application. 'CMU_WQuatNet.pth here' and 'AGORA_WQuatNet.pth here' for the full range angles or '300W_LP_WQuatNet.pth here' for narrow range angles.

Citing

@article{algabri2025wquatnet,
  title={WQuatNet: Wide range quaternion-based head pose estimation},
  author={Algabri, Redhwan and Shin, Hyunsoo and Abdu, Ahmed and Bae, Ji-Hun and Lee, Sungon},
  journal={Journal of King Saud University Computer and Information Sciences},
  volume={37},
  number={3},
  pages={24},
  year={2025},
  publisher={Springer}
}