Unsupervised Domain Adaptation Advances

The field of unsupervised domain adaptation is rapidly advancing, with a focus on developing innovative methods to bridge the gap between different domains. Researchers are exploring new architectures and techniques to enable knowledge transfer between distinct modalities, such as images and depth data. Notable progress has been made in addressing the challenges of heterogeneous-modal unsupervised domain adaptation, multi-modality domain adaptation, and simulation-to-reality point cloud recognition. Some noteworthy papers in this area include:

  • Heterogeneous-Modal Unsupervised Domain Adaptation via Latent Space Bridging, which proposes a novel framework for knowledge transfer between completely different modalities.
  • TITAN: Query-Token based Domain Adaptive Adversarial Learning, which introduces a target-based iterative query-token adversarial network for source-free domain adaptive object detection.

Sources

Heterogeneous-Modal Unsupervised Domain Adaptation via Latent Space Bridging

Self-Paced Collaborative and Adversarial Network for Unsupervised Domain Adaptation

Progressive Modality Cooperation for Multi-Modality Domain Adaptation

Boosting Domain Generalized and Adaptive Detection with Diffusion Models: Fitness, Generalization, and Transferability

Topology-Aware Modeling for Unsupervised Simulation-to-Reality Point Cloud Recognition

DuET: Dual Incremental Object Detection via Exemplar-Free Task Arithmetic

TITAN: Query-Token based Domain Adaptive Adversarial Learning

Built with on top of