Transflow: Transformer as flow learner

Y Lu, Q Wang, S Ma, T Geng… - Proceedings of the …, 2023 - openaccess.thecvf.com
Optical flow is an indispensable building block for various important computer vision tasks,
including motion estimation, object tracking, and disparity measurement. In this work, we …

Bridging the domain gap: Self-supervised 3d scene understanding with foundation models

Z Chen, L Jing, Y Li, B Li - Advances in Neural Information …, 2024 - proceedings.neurips.cc
Foundation models have achieved remarkable results in 2D and language tasks like image
segmentation, object detection, and visual-language understanding. However, their …

Sc-depthv3: Robust self-supervised monocular depth estimation for dynamic scenes

L Sun, JW Bian, H Zhan, W Yin, I Reid… - IEEE Transactions on …, 2023 - ieeexplore.ieee.org
Self-supervised monocular depth estimation has shown impressive results in static scenes. It
relies on the multi-view consistency assumption for training networks, however, that is …

Learning to fuse monocular and multi-view cues for multi-frame depth estimation in dynamic scenes

R Li, D Gong, W Yin, H Chen, Y Zhu… - Proceedings of the …, 2023 - openaccess.thecvf.com
Multi-frame depth estimation generally achieves high accuracy relying on the multi-view
geometric consistency. When applied in dynamic scenes, eg, autonomous driving, this …

Sqldepth: Generalizable self-supervised fine-structured monocular depth estimation

Y Wang, Y Liang, H Xu, S Jiao, H Yu - Proceedings of the AAAI …, 2024 - ojs.aaai.org
Recently, self-supervised monocular depth estimation has gained popularity with numerous
applications in autonomous driving and robotics. However, existing solutions primarily seek …

Naruto: Neural active reconstruction from uncertain target observations

Z Feng, H Zhan, Z Chen, Q Yan, X Xu… - Proceedings of the …, 2024 - openaccess.thecvf.com
We present NARUTO a neural active reconstruction system that combines a hybrid neural
representation with uncertainty learning enabling high-fidelity surface reconstruction. Our …

CVRecon: Rethinking 3d geometric feature learning for neural reconstruction

Z Feng, L Yang, P Guo, B Li - Proceedings of the IEEE/CVF …, 2023 - openaccess.thecvf.com
Recent advances in neural reconstruction using posed image sequences have made
remarkable progress. However, due to the lack of depth information, existing volumetric …

Know Your Neighbors: Improving Single-View Reconstruction via Spatial Vision-Language Reasoning

R Li, T Fischer, M Segu, M Pollefeys… - Proceedings of the …, 2024 - openaccess.thecvf.com
Recovering the 3D scene geometry from a single view is a fundamental yet ill-posed
problem in computer vision. While classical depth estimation methods infer only a 2.5 D …

[PDF][PDF] Deflowslam: Self-supervised scene motion decomposition for dynamic dense slam

W Ye, X Yu, X Lan, Y Ming, J Li, H Bao… - arXiv preprint arXiv …, 2022 - zju3dv.github.io
We present a novel dual-flow representation of scene motion that decomposes the optical
flow into a static flow field caused by the camera motion and another dynamic flow field …

Ds-depth: Dynamic and static depth estimation via a fusion cost volume

X Miao, Y Bai, H Duan, Y Huang, F Wan… - … on Circuits and …, 2023 - ieeexplore.ieee.org
Self-supervised monocular depth estimation methods typically rely on the reprojection error
to capture geometric relationships between successive frames in static environments …