Open Compound Domain Adaptation

By Ziwei Liu, Zhongqi Miao, Xingang Pan, Xiaohang Zhan, Dahua Lin, Stella X. Yu, Boqing Gong. 2019

144 citations recorded by the source corpus.

A typical domain adaptation approach is to adapt models trained on the annotated data in a source domain (e.g., sunny weather) for achieving high performance on the test data in a target domain (e.g., rainy weather). Whether the target contains a single homogeneous domain or multiple heterogeneous domains, existing works always assume that there exist clear distinctions between the domains, which is often not true in practice (e.g., changes in weather). We study an open compound domain adaptation (OCDA) problem, in which the target is a compound of multiple homogeneous domains without domain labels, reflecting realistic data collection from mixed and novel situations. We propose a new approach based on two technical insights into OCDA: 1) a curriculum domain adaptation strategy to bootstrap generalization across domains in a data-driven self-organizing fashion and 2) a memory module to increase the model's agility towards novel domains. Our experiments on digit classification, facial expression recognition, semantic segmentation, and reinforcement learning demonstrate the effectiveness of our approach.

Subfields: Artificial Intelligence, Computer Vision and Pattern Recognition, Radiology, Nuclear Medicine and Imaging, Representation & Self-Supervised Learning

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

In Radiology, Nuclear Medicine and Imaging

In Representation & Self-Supervised Learning

Citations in the corpus

Rescaling Egocentric Vision: Collection, Pipeline and Challenges for EPIC-KITCHENS-100 (2020, background): “[54], this is a more realistic scenario resulting from unconstrained data collection.”
Adapting Object Detectors with Conditional Domain Normalization (2020, background): “Domain Adaptation generalizes a model across different domains, and it has been extensively explored in various tasks, spanning from image classification [2, 40, 24, 27, 23], semantic segmentation [15, 39, 36] to reinforcement learning [38, 28, 20].”
Generalized Out-of-Distribution Detection: A Survey (2021, unknown): “covariate shift during testing without any semantic shift, and requires classifiers to make accurate predictions into the same set of classes [92].”
Transformer-Based Visual Segmentation: A Survey (2023, background): “• Domain Adaption: Unsupervised Domain Adaptation (UDA) aims at adapting the network trained with source (synthetic) domain into target (real) domain [45], [289] without access to target labels.”
Domain Generalization: A Survey (2021, background): “DG, such as single- [150] and multi-source [60] DA, and heterogeneous DA [17], [156], [157], [158].”
Guided Curriculum Model Adaptation and Uncertainty-Aware Evaluation for Semantic Nighttime Image Segmentation (2019, methodology): “lignment between domains. The general concept of curriculum learning has been successfully applied to domain adaptation by ordering tasks [46], target-domain pixels [47], or domains [10], [11], [34], [48]. Our method belongs to the last group. Cross-domain correspondences as guidance have only been ”
Overcoming Negative Transfer: A Survey (2020, methodology): “Liu et al. [54] showed a more realistic scenario called open compound that the target is a compound of multiple homogeneous domains without domain labels, reﬂecting realistic data collection from mixed and novel situations, and they adopted a curriculum domain adaptation strategy to bootstrap…”
Distribution-Balanced Loss for Multi-Label Classification in Long-Tailed Datasets (2020, background): “While providing a clean setting for various studies, this conventional setting conceals a number of complexities that often arise in real-world applications [27,26,16,14].”
Discover, Hallucinate, and Adapt: Open Compound Domain Adaptation for Semantic Segmentation (2021, methodology): “• Testing We follow the conventional inference setup [23].”