Martyniuk_DAD-3DHeads_A_Large-Scale_Dense_Accurate_and_Diverse_Dataset_for_3D_CVPR_2022_paper.pdf (3.61 MB)
DAD-3DHeads: a large-scale dense, accurate and diverse dataset for 3D head alignment from a single image
conference contribution
posted on 2023-06-10, 04:26 authored by Tetiana Martyniuk, Orest Kupyn, Yana Kurlyak, Igor Krashenyi, Jiri Matas, Viktoriia SharmanskaViktoriia SharmanskaWe present DAD-3DHeads, a dense and diverse large-scale dataset, and a robust model for 3D Dense Head Alignment in-the-wild. It contains annotations of over 3.5K landmarks that accurately represent 3D head shape compared to the ground-truth scans. The data-driven model, DAD-3DNet, trained on our dataset, learns shape, expression, and pose parameters, and performs 3D reconstruction of a FLAME mesh. The model also incorporates a landmark prediction branch to take advantage of rich supervision and co-training of multiple related tasks. Experimentally, DAD- 3DNet outperforms or is comparable to the state-of-the-art models in (i) 3D Head Pose Estimation on AFLW2000-3D and BIWI, (ii) 3D Face Shape Reconstruction on NoW and Feng, and (iii) 3D Dense Head Alignment and 3D Landmarks Estimation on DAD-3DHeads dataset. Finally, diversity of DAD-3DHeads in camera angles, facial expressions, and occlusions enables a benchmark to study in-the-wild generalization and robustness to distribution shifts. The dataset webpage is https://p.farm/research/dad-3dheads.
History
Publication status
- Published
File Version
- Accepted version
Journal
Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR), 2022Publisher
IEEEExternal DOI
Page range
20942-20952Event name
IEEE Conference on Computer Vision and Pattern Recognition (CVPR)Event location
New Orleans, USAEvent type
conferenceEvent date
19th - 24th June 2022ISBN
9781665469463Department affiliated with
- Informatics Publications
Full text available
- Yes
Peer reviewed?
- No
Legacy Posted Date
2022-08-09First Open Access (FOA) Date
2022-11-30First Compliant Deposit (FCD) Date
2022-08-08Usage metrics
Categories
No categories selectedKeywords
Licence
Exports
RefWorks
BibTeX
Ref. manager
Endnote
DataCite
NLM
DC