DESC: Domain Adaptation for Depth Estimation via Semantic Consistency

Lopez-Rodriguez, Adrian; Mikolajczyk, Krystian

doi:10.48550/arxiv.2009.01579

Cited by 4 publications

(6 citation statements)

References 47 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…In [32,54,59] source domain data is replaced with a model pre-trained on the source. [36] uses low-level edge features to enforce consistency in UDA for monocular depth estimation. In Fewshot UDA (FUDA) [26,60], only a few examples per class are labeled in the source domain, while the rest are unlabeled.…”

Section: Related Workmentioning

confidence: 99%

Unsupervised Domain Generalization by Learning a Bridge Across Domains

Harary¹,

Schwartz²,

Arbelle³

et al. 2021

Preprint

View full text Add to dashboard Cite

The ability to generalize learned representations across significantly different visual domains, such as between real photos, clipart, paintings, and sketches, is a fundamental capacity of the human visual system. In this paper, different from most cross-domain works that utilize some (or full) source domain supervision, we approach a relatively new and very practical Unsupervised Domain Generalization (UDG) setup of having no training supervision in neither source nor target domains. Our approach is based on selfsupervised learning of a Bridge Across Domains (BrAD)an auxiliary bridge domain accompanied by a set of semantics preserving visual (image-to-image) mappings to BrAD from each of the training domains. The BrAD and mappings to it are learned jointly (end-to-end) with a contrastive selfsupervised representation model that semantically aligns each of the domains to its BrAD-projection, and hence implicitly drives all the domains (seen or unseen) to semantically align to each other. In this work, we show how using an edge-regularized BrAD our approach achieves significant gains across multiple benchmarks and a range of tasks, including UDG, Few-shot UDA, and unsupervised generalization across multi-domain datasets (including generalization to unseen domains and classes).

show abstract

Section: Related Workmentioning

confidence: 99%

Unsupervised Domain Generalization by Learning a Bridge Across Domains

Harary¹,

Schwartz²,

Arbelle³

et al. 2021

Preprint

View full text Add to dashboard Cite

show abstract

“…The first category mainly covers the conventional methods including discrepancy measures such as MMD [50,22] and CMD [99], geodesic flow kernel [28], sub-space alignment [19], asymmetric metric learning [41], etc. The general idea of the second category is to align source and target domains at different representation levels, including: (1) input image-level alignment [5,36] using imageto-image translation methods such as CycleGAN [109], or statistics matching [1]; (2) internal feature-level alignment based on feature-level domain adversarial learning [83,51,105]; and (3) output-space alignment [81,85,52] typically by an adversarial module. For the third category, self-supervised learning based domain adaptation methods [23] achieve great progress, in which simple auxiliary tasks generated automatically from unlabeled data are utilized to train feature representations, such as rotation prediction [24], flip prediction [90], patch location prediction [90], etc.…”

Section: Domain Adaptationmentioning

confidence: 99%

“…However, much less attention has been paid to domain adaptation for low-level tasks. There are several works investigating the domainadaptive depth estimation task, including geometry-aware alignment [105], semantic-level consistency [52] and imagelevel translation [106,2].…”

Section: Domain Adaptationmentioning

confidence: 99%

AdaStereo: An Efficient Domain-Adaptive Stereo Matching Approach

Song¹,

Yang²,

Zhu³

et al. 2021

Preprint

View full text Add to dashboard Cite

Recently, records on stereo matching benchmarks are constantly broken by end-to-end disparity networks. However, the domain adaptation ability of these deep models is quite limited. Addressing such problem, we present a novel domain-adaptive approach called AdaStereo that aims to align multi-level representations for deep stereo matching networks. Compared to previous methods, our AdaStereo realizes a more standard, complete and effective domain adaptation pipeline. Firstly, we propose a non-adversarial progressive color transfer algorithm for input image-level alignment. Secondly, we design an efficient parameter-free cost normalization layer for internal feature-level alignment. Lastly, a highly related auxiliary task, self-supervised occlusionaware reconstruction is presented to narrow the gaps in output space. We perform intensive ablation studies and breakdown comparisons to validate the effectiveness of each proposed module. With no extra inference overhead and only a slight increase in training complexity, our AdaStereo models achieve state-of-the-art cross-domain performance on multiple benchmarks, including KITTI, Middlebury, ETH3D and DrivingStereo, even outperforming some state-of-the-art disparity networks finetuned with target-domain ground-truths. Moreover, based on two additional evaluation metrics, the superiority of our domain-adaptive stereo matching pipeline is further uncovered from more perspectives. Finally, we demonstrate that our method is robust to various domain adaptation settings, and can be easily integrated into quick adaptation application scenarios and real-world deployments.

show abstract

“…Unsupervised domain adaptation attempts to overcome these limitations. However, the vast majority of proposed approaches focus on sim-to-real domain adaptation mostly in an offline manner [12,23], i.e., a directed knowledge transfer without the need to avoid catastrophic forgetting and with access to Fig. 1.…”

Section: Introductionmentioning

confidence: 99%

Continual SLAM: Beyond Lifelong Simultaneous Localization and Mapping Through Continual Learning

Vödisch

Cattaneo

Burgard

et al. 2023

Springer Proceedings in Advanced Robotics

View full text Add to dashboard Cite

Visual odometry is a fundamental task for many applications on mobile devices and robotic platforms. Since such applications are oftentimes not limited to predefined target domains and learning-based vision systems are known to generalize poorly to unseen environments, methods for continual adaptation during inference time are of significant interest. In this work, we introduce CoVIO for online continual learning of visual-inertial odometry. CoVIO effectively adapts to new domains while mitigating catastrophic forgetting by exploiting experience replay. In particular, we propose a novel sampling strategy to maximize image diversity in a fixed-size replay buffer that targets the limited storage capacity of embedded devices. We further provide an asynchronous version that decouples the odometry estimation from the network weight update step enabling continuous inference in real time. We extensively evaluate CoVIO on various real-world datasets demonstrating that it successfully adapts to new domains while outperforming previous methods. The code of our work is publicly available at http://continual-slam.cs.uni-freiburg.de.

show abstract

DESC: Domain Adaptation for Depth Estimation via Semantic Consistency

Cited by 4 publications

References 47 publications

Unsupervised Domain Generalization by Learning a Bridge Across Domains

Unsupervised Domain Generalization by Learning a Bridge Across Domains

AdaStereo: An Efficient Domain-Adaptive Stereo Matching Approach

Continual SLAM: Beyond Lifelong Simultaneous Localization and Mapping Through Continual Learning

Contact Info

Product

Resources

About