TasksSotADatasetsPapersMethodsSubmitAbout
Papers With Code 2

A community resource for machine learning research: papers, code, benchmarks, and state-of-the-art results.

Explore

Notable BenchmarksAll SotADatasetsPapersMethods

Community

Submit ResultsAbout

Data sourced from the PWC Archive (CC-BY-SA 4.0). Built by the community, for the community.

Papers/On the power of data augmentation for head pose estimation

On the power of data augmentation for head pose estimation

Michael Welter

2024-07-07Face AlignmentData AugmentationPose EstimationHead Pose Estimation
PaperPDFCode(official)

Abstract

Deep learning has been impressively successful in the last decade in predicting human head poses from monocular images. However, for in-the-wild inputs the research community relies predominantly on a single training set, 300W-LP, of semisynthetic nature without many alternatives. This paper focuses on gradual extension and improvement of the data to explore the performance achievable with augmentation and synthesis strategies further. Modeling-wise a novel multitask head/loss design which includes uncertainty estimation is proposed. Overall, the thus obtained models are small, efficient, suitable for full 6 DoF pose estimation, and exhibit very competitive accuracy.

Results

TaskDatasetMetricValueModel
Pose EstimationAFLW2000Geodesic Error (GE)5.23OpNet
Pose EstimationAFLW2000MAE3.15OpNet
Pose EstimationBIWIGeodesic Error (GE)7.01OpNet
Pose EstimationBIWIGeodesic Error - aligned (GE)4.72OpNet
Pose EstimationBIWIMAE (trained with other data)3.57OpNet
Pose EstimationBIWIMAE-aligned (trained with other data)2.65OpNet
3DAFLW2000Geodesic Error (GE)5.23OpNet
3DAFLW2000MAE3.15OpNet
3DBIWIGeodesic Error (GE)7.01OpNet
3DBIWIGeodesic Error - aligned (GE)4.72OpNet
3DBIWIMAE (trained with other data)3.57OpNet
3DBIWIMAE-aligned (trained with other data)2.65OpNet
1 Image, 2*2 StitchiAFLW2000Geodesic Error (GE)5.23OpNet
1 Image, 2*2 StitchiAFLW2000MAE3.15OpNet
1 Image, 2*2 StitchiBIWIGeodesic Error (GE)7.01OpNet
1 Image, 2*2 StitchiBIWIGeodesic Error - aligned (GE)4.72OpNet
1 Image, 2*2 StitchiBIWIMAE (trained with other data)3.57OpNet
1 Image, 2*2 StitchiBIWIMAE-aligned (trained with other data)2.65OpNet

Related Papers

Overview of the TalentCLEF 2025: Skill and Job Title Intelligence for Human Capital Management2025-07-17Pixel Perfect MegaMed: A Megapixel-Scale Vision-Language Foundation Model for Generating High Resolution Medical Images2025-07-17$π^3$: Scalable Permutation-Equivariant Visual Geometry Learning2025-07-17Revisiting Reliability in the Reasoning-based Pose Estimation Benchmark2025-07-17DINO-VO: A Feature-based Visual Odometry Leveraging a Visual Foundation Model2025-07-17From Neck to Head: Bio-Impedance Sensing for Head Pose Estimation2025-07-17AthleticsPose: Authentic Sports Motion Dataset on Athletic Field and Evaluation of Monocular 3D Pose Estimation Ability2025-07-17Similarity-Guided Diffusion for Contrastive Sequential Recommendation2025-07-16