To facilitate the zero-shot classification task focusing on Mars scenes, we reclassified the image categories within the collected public datasets. By reviewing the two versions of the Mars surface image datasets, MSL (v1) and MSL (v2), we identified overlapping categories and inconsistencies in size and color across images. To resolve these issues, we standardized the size and color of the images across the datasets and redefined the image categories specific to the MSC. We then formed an annotation team to conduct visual image annotation on the revised ZSMSC dataset, now named ZSMars.
If it is helpful for your work, please cite this paper:
@ARTICLE{10699382,
author={Tan, Xiaomeng and Xi, Bobo and Xu, Haitao and Li, Jiaojiao and Li, Yunsong and Xue, Changbin and Chanussot, Jocelyn},
journal={IEEE Transactions on Geoscience and Remote Sensing},
title={A Lightweight Framework with Knowledge Distillation for Zero-Shot Mars Scene Classification},
year={2024},
volume={},
number={},
pages={1-1},
keywords={Mars;Visualization;Semantics;Feature extraction;Scene classification;Image recognition;Microwave integrated circuits;Accuracy;Transformers;Data models;Mars scene classification;zero-shot learning;knowledge distillation;lightweight model},
doi={10.1109/TGRS.2024.3470526}}