awesome-domain-adaptation is a curated collection of resources about domain adaptation research and applications.
The project addresses the need for researchers and practitioners to discover relevant work in domain adaptation, a field concerned with training models on one domain and applying them effectively to different domains. It organizes papers, code, benchmarks, and other resources into a structured taxonomy covering the breadth of domain adaptation approaches and problem settings.
Developers and researchers working on transfer learning problems should use this collection to survey the landscape of domain adaptation methods and find relevant papers for their specific scenario. The taxonomy distinguishes between numerous problem formulations including unsupervised, semi-supervised, weakly-supervised, zero-shot, few-shot, partial, open-set, multi-source, and federated domain adaptation, as well as related paradigms like domain generalization and domain randomization. It also covers domain adaptation applications across object detection, semantic segmentation, person re-identification, sim-to-real transfer, video analysis, medical imaging, depth estimation, 3D vision, and remote sensing. This breadth makes it useful for identifying both methodological approaches and application-specific solutions.
The project functions as a community-maintained reference rather than an actively developed tool, serving primarily as an organizational index for the domain adaptation literature and related resources including benchmarks, libraries, and educational materials.
