Exploiting domain transferability for collaborative inter-level domain adaptive object detection

作者：

Highlights：

• Reducing the domain gap of individual levels in the two-stage detector pipeline.

• Aggregating multi-scale image-level features for transferability estimation.

• Effective reduction of the domain gap with selected adaptive-proper region proposals.

• Preventing the negative effect caused by inaccurate target proposals during training.

• Experiments on various domain adaptation scenarios of driving scenes.

摘要

•Reducing the domain gap of individual levels in the two-stage detector pipeline.•Aggregating multi-scale image-level features for transferability estimation.•Effective reduction of the domain gap with selected adaptive-proper region proposals.•Preventing the negative effect caused by inaccurate target proposals during training.•Experiments on various domain adaptation scenarios of driving scenes.

论文关键词：Domain adaptation,Object detection,Domain transferability,Domain shift

论文评审过程：Received 11 March 2022, Revised 7 May 2022, Accepted 28 May 2022, Available online 3 June 2022, Version of Record 10 June 2022.

论文官网地址：https://doi.org/10.1016/j.eswa.2022.117697