Abstract:Cross domain object detection learns an object detector for an unlabeled target domain by transferring knowledge from an annotated source domain. Promising results have been achieved via Mean Teacher, however, pseudo labeling which is the bottleneck of mutual learning remains to be further explored. In this study, we find that confidence misalignment of the predictions, including category-level overconfidence, instance-level task confidence inconsistency, and image-level confidence misfocusing, leading to the injection of noisy pseudo label in the training process, will bring suboptimal performance on the target domain. To tackle this issue, we present a novel general framework termed Multi-Granularity Confidence Alignment Mean Teacher (MGCAMT) for cross domain object detection, which alleviates confidence misalignment across category-, instance-, and image-levels simultaneously to obtain high quality pseudo supervision for better teacher-student learning. Specifically, to align confidence with accuracy at category level, we propose Classification Confidence Alignment (CCA) to model category uncertainty based on Evidential Deep Learning (EDL) and filter out the category incorrect labels via an uncertainty-aware selection strategy. Furthermore, to mitigate the instance-level misalignment between classification and localization, we design Task Confidence Alignment (TCA) to enhance the interaction between the two task branches and allow each classification feature to adaptively locate the optimal feature for the regression. Finally, we develop imagery Focusing Confidence Alignment (FCA) adopting another way of pseudo label learning, i.e., we use the original outputs from the Mean Teacher network for supervised learning without label assignment to concentrate on holistic information in the target image. These three procedures benefit from each other from a cooperative learning perspective.

A Teacher-Student Approach to Cross-Domain Transfer Learning with Multi-level Attention

Transfer Learning and Deep Domain Adaptation

Sharing, Teaching and Aligning: Knowledgeable Transfer Learning for Cross-Lingual Machine Reading Comprehension

Collaborative Teaching with Attention Distillation for Multiple Cross-Domain Few-Shot Learning

Interactive Attention Transfer Network for Cross-Domain Sentiment Classification

Three Heads Are Better Than One: Improving Cross-Domain NER with Progressive Decomposed Network

Versatile Teacher: A class-aware teacher–student framework for cross-domain adaptation

Cross-Domain Adaptive Teacher for Object Detection

Cross Domain Object Detection via Multi-Granularity Confidence Alignment based Mean Teacher

From multi-label learning to cross-domain transfer: a model-agnostic approach

Multilevel Knowledge Transfer for Cross-Domain Object Detection

Joint Attention-Driven Domain Fusion and Noise-Tolerant Learning for Multi-Source Domain Adaptation

Multi-Level Domain Adaptive Learning For Cross-Domain Detection

Exploring Discriminative Word-Level Domain Contexts for Multi-Domain Neural Machine Translation

Adaptive and Dynamic Knowledge Transfer in Multi-task Learning with Attention Networks.

Cross-domain NER under a Divide-and-Transfer Paradigm

Versatile Teacher: A Class-aware Teacher-student Framework for Cross-domain Adaptation

Translation as Cross-Domain Knowledge - Attention Augmentation for Unsupervised Cross-Domain Segmenting and Labeling Tasks.

Single-/Multi-Source Cross-Lingual NER via Teacher-Student Learning on Unlabeled Data in Target Language

Transductive transfer broad learning for cross-domain information exploration and multigrade soft sensor application

Towards Understanding Transfer Learning Algorithms Using Meta Transfer Features