In order to truly make progress in the problem of predicting pose from image intensities, real datasets which contain precise pose annotations, numerous identities, different lighting conditions, all of this across large poses occur. Most of the HPE models are evaluated using publicly available datasets. These datasets significantly evolved during the last years, especially in terms of complexity of environmental conditions. Head pose datasets can be categorized by different aspects, such as imaging characteristics (e.g. RGB, grayscale, depth, infrared images), data diversity (e.g. head pose angles ranges, or number of degrees of freedom), acquisition scenario (e.g. in laboratory vs in-the-wild), annotation type, and annotation technique.
This list contains the datasets used in the literature for the head pose estimation task:
- 300W-LP: https://www.tensorflow.org/datasets/catalog/the300w\_lp
paper: https://arxiv.org/abs/1511.07212 - AFLW: https://www.tugraz.at/institute/icg/research/team-bischof/lrs/downloads/aflw/
paper: https://ieeexplore.ieee.org/abstract/document/6130513 - AFLW2000-3D: https://www.tensorflow.org/datasets/catalog/aflw2k3d
paper: https://arxiv.org/abs/1511.07212 - AFW: https://ibug.doc.ic.ac.uk/resources/facial-point-annotations/
paper: https://ieeexplore.ieee.org/abstract/document/6248014 - AISL: http://www.aisl.cs.tut.ac.jp/dataset\_head\_orientation.html
paper: https://ieeexplore.ieee.org/abstract/document/7380830 - AutoPOSE: https://autopose.dfki.de/
paper: https://www.researchgate.net/profile/Mohamed-Selim-12/publication/340042088_AutoPOSE_Large-scale_Automotive_Driver_Head_Pose_and_Gaze_Dataset_with_Deep_Head_Orientation_Baseline/links/5e7877b3a6fdcccd62192067/AutoPOSE-Large-scale-Automotive-Driver-Head-Pose-and-Gaze-Dataset-with-Deep-Head-Orientation-Baseline.pdf - BioVid Heat Pain: https://www.iikt.ovgu.de/BioVid.html
paper: https://ieeexplore.ieee.org/abstract/document/6617456 - BIWI Kinect: https://www.kaggle.com/kmader/biwi-kinect-head-pose-database
paper: https://link.springer.com/chapter/10.1007/978-3-642-23123-0_11 - BJUT-3D: http://www.bjpu.edu.cn/sci/multimedia/mul-lab/3dface/facedatabase.htm
paper: https://crad.ict.ac.cn/EN/abstract/abstract1023.shtml - Bosphorus: http://bosphorus.ee.boun.edu.tr/default.aspx
paper: https://link.springer.com/chapter/10.1007/978-3-540-89991-4_6 - BU: https://www.cs.bu.edu/groups/ivc/HeadTracking/Home.html
paper: https://ieeexplore.ieee.org/document/845375 - CAS-PEAL: http://www.jdl.ac.cn/peal
paper: https://ieeexplore.ieee.org/abstract/document/4404053 - CAVE (Columbia Gaze): https://www.cs.columbia.edu/CAVE/databases/columbia\_gaze/
paper: https://dl.acm.org/doi/abs/10.1145/2501988.2501994 - CCNU
paper: https://www.sciencedirect.com/science/article/pii/S0925231215010413 - CMU Multi-Pie: https://www.cs.cmu.edu/afs/cs/project/PIE/MultiPie/Multi-Pie/Home.html
paper: https://www.sciencedirect.com/science/article/pii/S0262885609001711 - CMU Panoptic: http://domedb.perception.cs.cmu.edu/
paper: https://openaccess.thecvf.com/content_iccv_2015/papers/Joo_Panoptic_Studio_A_ICCV_2015_paper.pdf
Database processed for head pose: Ascend-Research/HeadPoseEstimation-WHENet#13 - CMU-Pie: https://www.ri.cmu.edu/project/pie-database/
paper: https://www.ri.cmu.edu/pub_files/pub2/sim_terence_2001_1/sim_terence_2001_1.pdf - Dali3DHP
paper: https://ieeexplore.ieee.org/document/6977105 - DD-Pose: https://dd-pose-dataset.tudelft.nl/eval/
paper: https://ieeexplore.ieee.org/abstract/document/8814103 - DriveAHead: https://cvhci.anthropomatik.kit.edu/~aschwarz/driveahead/
paper: https://openaccess.thecvf.com/content_cvpr_2017_workshops/w13/html/Schwarz_DriveAHead_-_A_CVPR_2017_paper.html - ETH: https://data.vision.ee.ethz.ch/cvl/vision2/datasets/headposeCVPR08/
paper: https://ieeexplore.ieee.org/document/4587807 - FacePix: https://cubic.asu.edu/content/facepix-database
paper: https://ieeexplore.ieee.org/document/1415348 - GI4E-HP: http://www.unavarra.es/gi4e/databases?languageId=1
paper: https://dl.acm.org/doi/10.5555/2951132.2951428 - GOTCHA-I: https://gotchaproject.github.io/
paper: https://link.springer.com/chapter/10.1007/978-981-15-4825-3_17 - ICT-3DHP: http://multicomp.cs.cmu.edu/resources/ict-3d-headpose-database-2/
paper: https://ieeexplore.ieee.org/document/6247980 - IDIAP Head Pose: https://www.idiap.ch/en/dataset/headpose
paper: http://publications.idiap.ch/index.php/publications/show/349 - M2FPA: https://pp2li.github.io/M2FPA-dataset/
paper: https://openaccess.thecvf.com/content_ICCV_2019/papers/Li_M2FPA_A_Multi-Yaw_Multi-Pitch_High-Quality_Dataset_and_Benchmark_for_Facial_ICCV_2019_paper.pdf - McGill: https://sites.google.com/site/meltemdemirkus/mcgill-unconstrained-face-video-database
paper: https://link.springer.com/article/10.1007\%2Fs11042-012-1352-1 - MDM Corpus: https://ecs.utdallas.edu/research/researchlabs/msp-lab/MDM.html
paper: https://ieeexplore.ieee.org/abstract/document/9507390 - MTFL: http://mmlab.ie.cuhk.edu.hk/projects/TCDCN.html
paper: http://personal.ie.cuhk.edu.hk/~ccloy/files/eccv_2014_deepfacealign.pdf - Pandora: https://aimagelab.ing.unimore.it/pandora/
paper: https://openaccess.thecvf.com/content_cvpr_2017/papers/Borghi_POSEidon_Face-From-Depth_for_CVPR_2017_paper.pdf - Pointing'04: http://crowley-coutaz.fr/Head\%20Pose\%20Image\%20Database.html
paper: https://citeseerx.ist.psu.edu/viewdoc/download?doi=10.1.1.381.3419&rep=rep1&type=pdf - SASE: https://icv.tuit.ut.ee/databases/
paper: https://ieeexplore.ieee.org/document/7961825 - SyLaHP: https://www.iikt.ovgu.de/LmHeadPoseEstBench.html
paper: https://ieeexplore.ieee.org/abstract/document/8297015 - SynHead: http://www.tnt.uni-hannover.de/papers/view\_paper.php?id=1419
paper: https://openaccess.thecvf.com/content_cvpr_2017/html/Gu_Dynamic_Facial_Analysis_CVPR_2017_paper.html - Synthetic: https://liangwei-bit.github.io/web/project/icip16\_headpose/
paper: https://ieeexplore.ieee.org/document/7532566 - Taiwan RoboticsLab: http://robotics.csie.ncku.edu.tw/Databases/FaceDetect\_PoseEstimate.htm
paper: https://www.researchgate.net/publication/222079557_A_view-based_statistical_system_for_multi-view_face_detection_and_pose_estimation - UbiPose: https://www.idiap.ch/en/dataset/ubipose
paper: https://www.idiap.ch/~odobez/publications/YuFunesOdobez-PAMI2018.pdf - UET-Headpose
paper: https://arxiv.org/abs/2111.07039?context=cs.HC - UMD Faces: http://umdfaces.io/
paper: https://arxiv.org/pdf/1611.01484.pdf - VGGFace2: https://github.com/ox-vgg/vgg\_face2
paper: https://ieeexplore.ieee.org/abstract/document/8373813
For more details about each dataset, or acquisition methods refer to https://amslaurea.unibo.it/25279/. In this thesis you can also find details on methods for head pose estimation, evaluation metrics, evaluation pipelines and results obtained on commonly used datasets.
Article 📑 Asperti, A., Filippini, D. Deep Learning for Head Pose Estimation: A Survey. SN COMPUT. SCI. 4, 349 (2023). https://doi.org/10.1007/s42979-023-01796-z
UPDATES
New Datasets
- 2DHeadPose: https://github.com/youngnuaa/2DHeadPose
paper: https://doi.org/10.1016/j.neunet.2022.12.021 - AGORA-HPE: https://github.com/hnuzhy/DirectMHP?tab=readme-ov-file#agora-hpe-dataset
[Daniele Filippini]