Incremental Adversarial Domain Adaptation for Continually Changing Environments

By Markus Wulfmeier, Alex Bewley, Ingmar Posner. 2017

138 citations recorded by the source corpus.

Continuous appearance shifts such as changes in weather and lighting conditions can impact the performance of deployed machine learning models. While unsupervised domain adaptation aims to address this challenge, current approaches do not utilise the continuity of the occurring shifts. In particular, many robotics applications exhibit these conditions and thus facilitate the potential to incrementally adapt a learnt model over minor shifts which integrate to massive differences over time. Our work presents an adversarial approach for lifelong, incremental domain adaptation which benefits from unsupervised alignment to a series of intermediate domains which successively diverge from the labelled source domain. We empirically demonstrate that our incremental approach improves handling of large appearance changes, e.g. day to night, on a traversable-path segmentation task compared with a direct, single alignment step approach. Furthermore, by approximating the feature distribution for the source domain with a generative adversarial network, the deployment module can be rendered fully independent of retaining potentially large amounts of the related source training data for only a minor reduction in performance.

Subfields: Artificial Intelligence, Computer Vision and Pattern Recognition

How researchers see this paper

The researchers whose trusted networks rank it highest, strongest advocate first. Every researcher gets a different order.

In Artificial Intelligence

In Computer Vision and Pattern Recognition

Citations in the corpus

WILDS: A Benchmark of in-the-Wild Distribution Shifts (2020, neutral): “These problems have manifested in many ways, including shifts induced by weather and lighting changes (Wulfmeier et al., 2018), location changes (Gupta et al., 2018), and the simulation-to-real-world gap (Sadeghi and Levine, 2017; Tobin et al., 2017).”
Continual Test-Time Domain Adaptation (2022, background): “Incremental adversarial domain adaptation (IADA) [63] adapts to continually changing domains by adversarially aligning source and target features.”
Gradient Regularized Contrastive Learning for Continual Domain Adaptation (2020, methodology): “The approaches (Mancini et al. 2019; Gong et al. 2019; Wulfmeier, Bewley, and Posner 2018; Hoffman et al. 2014) do not explicitly target the catastrophic forgetting problem, while the approaches (Bobu et al. 2018) and our approach target the catastrophic forgetting problem.”
Guided Curriculum Model Adaptation and Uncertainty-Aware Evaluation for Semantic Nighttime Image Segmentation (2019, background): “A major class of adaptation approaches, including [5,12,13,19,28,31,34,36,43,46], involves adversarial confusion or feature alignment between domains.”
DLOW: Domain Flow for Adaptation and Generalization (2018, background): “Domain adaptation aims to utilize a labeled source domain to learn a model that performs well on an unlabeled target domain [13, 18, 12, 55, 29, 3, 31, 16, 6, 61, 57].”
Continuously Indexed Domain Adaptation (2020, methodology): “, 2016), generative adversarial networks (Wulfmeier et al., 2018), and linear transform (Hoffman et al.”
Deep Unsupervised Domain Adaptation: A Review of Recent Advances and Perspectives (2022, unknown): “To accommodate continuously changing environments, we expect that frameworks employed would be source-free, involve low-cost training in mobile devices, and avoid catastrophic forgetting (Hoffman et al., 2014; Wulfmeier et al., 2018; Wu et al., 2019b; Mancini et al., 2019).”
Learning to Drive from Simulation without Real World Labels (2018, background): “However, feature based alignment typically requires some additional pairwise constraints, [29] or intermediate domains [34], [35], to bridge significant shifts in domain appearance.”
Continuous Domain Adaptation with Variational Domain-Agnostic Feature Replay (2020, background): “There are works, however, that aim to tackle the domain drift for steam data, i.e., a stream of continuously evolving domains, but their proposed methodologies still lack the consideration of task drift in the stream (Hoffman et al., 2014; Wulfmeier et al., 2018).”
A Survey of Unsupervised Deep Domain Adaptation (2018, methodology): “n the Office dataset, Table3), or simple to complex images (e.g., MNIST to SVHN, Table2). Others have been used in robotics for robot grasping [20], autonomous navigation [266], and lifelong learning [255], for semantic segmentation [40, 98, 102, 133, 153, 209, 239, 243, 291] including when addition”