Publication Details

Estimating Extreme 3D Image Rotations using Cascaded Attention

DEKEL Shay, KELLER Yosi and ČADÍK Martin. Estimating Extreme 3D Image Rotations using Cascaded Attention. In: Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR). Seattle: IEEE Computer Society, 2024, pp. 2588-2598. ISBN 979-8-3503-5301-3. Available from: https://cadik.posvete.cz/
Czech title
Odhad extrémních 3D rotací kamery mezi snímky pomocí cascaded attention
Type
conference paper
Language
english
Authors
Dekel Shay (BIU)
Keller Yosi, prof. MSc., Ph.D. (BIU)
Čadík Martin, doc. Ing., Ph.D. (DCGM FIT BUT)
URL
Keywords

camera orientation estimation, extreme rotation, 3D rotation, cascaded attention

Abstract

Estimating large, extreme inter-image rotations is critical for numerous computer vision domains involving images related by limited or non-overlapping fields of view. In this work, we propose an attention-based approach with a pipeline of novel algorithmic components. First, as rotation estimation pertains to image pairs, we introduce an inter-image distillation scheme using Decoders to improve embeddings. Second, whereas contemporary methods compute a 4D correlation volume (4DCV) encoding inter-image relationships, we propose an Encoder-based cross-attention approach between activation maps to compute an enhanced equivalent of the 4DCV. Finally, we present a cascaded Decoder-based technique for alternately refining the cross-attention and the rotation query. Our approach outperforms current state-of-the-art methods on extreme rotation estimation. We make our code publicly available.

Published
2024
Pages
2588-2598
Proceedings
Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR)
Conference
The IEEE/CVF Conference on Computer Vision and Pattern Recognition 2024, Seattle Convention Center, US
ISBN
979-8-3503-5301-3
Publisher
IEEE Computer Society
Place
Seattle, US
DOI
BibTeX
@INPROCEEDINGS{FITPUB13178,
   author = "Shay Dekel and Yosi Keller and Martin \v{C}ad\'{i}k",
   title = "Estimating Extreme 3D Image Rotations using Cascaded Attention",
   pages = "2588--2598",
   booktitle = "Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR)",
   year = 2024,
   location = "Seattle, US",
   publisher = "IEEE Computer Society",
   ISBN = "979-8-3503-5301-3",
   doi = "10.1109/CVPR52733.2024.00250",
   language = "english",
   url = "https://www.fit.vut.cz/research/publication/13178"
}
Back to top