A robust end-to-end deep learning framework for detecting Martian landforms with arbitrary orientations

作者：

Highlights：

• The framework can locate and identify different landforms on Mars in one pass via rotatable anchor-box based mechanism.

• A new efficient match strategy is proposed for anchoring default boxes to ground truth boxes.

• An autoencoder-based unsupervised pre-training operation is introduced.

• A new object detection dataset is constructed for landform detection tasks.

摘要

•The framework can locate and identify different landforms on Mars in one pass via rotatable anchor-box based mechanism.•A new efficient match strategy is proposed for anchoring default boxes to ground truth boxes.•An autoencoder-based unsupervised pre-training operation is introduced.•A new object detection dataset is constructed for landform detection tasks.

论文关键词：Deep learning,Object detection,Rotated single shot multibox detector,Match strategy,Autoencoder

论文评审过程：Received 30 August 2020, Revised 14 June 2021, Accepted 2 October 2021, Available online 5 October 2021, Version of Record 19 October 2021.

论文官网地址：https://doi.org/10.1016/j.knosys.2021.107562