MarsData-V2

This work is licensed under a Creative Commons Attribution .

We release MarsData-V2, a rock segmentation dataset of real Martian scenes for the training of deep networks, extended from our previously published MarsData [1]. The raw unlabeled RGB images of MarsData-V2 are from here, which were collected by a Mastcam camera of the Curiosity rover on Mars between August 2012 and November 2018. After sorting out images with an opportune shooting distance, labeling Mars rocks with fine-grained boundaries, and performing data augmentation referring to [2], we obtained 8390 labeled images with a resolution of 512 512, and divided them into 5040 images for training, 1680 for validation and 1670 for testing.

We show 8 samples, including 4 samples in MarsData and 4 new ones in MarsData-V2, where the Martian rocks with varying shapes, sizes, textures and colors are labeled with fine-grained annotations.

4 samples in MarsData:

4 new samples in MarsData-V2:

Limited by the file size, we temporarily release 100 samples in the train, validation and test set of MarsData-V2 on github, respectively. The whole data can be found by the following way:

[1]: IEEE DataPort; <!-- [2]: Baidu Cloud with passcode: rock

[3]: Our Lab filestation --> [2]: If you do not acess to IEEE DataPort, please contact us by email(alexcapshow@cust.edu.cn or meibaoyao@jlu.edu.cn) and sign the data license aggrement to get the dataset .

If you use MarsDataV2 for your research, please cite all the following papers and data

``` @article{liu2023rockformer, title={RockFormer: A U-Shaped Transformer Network for Martian Rock Segmentation}, author={Liu, Haiqiang and Yao, Meibao and Xiao, Xueming and Xiong, Yonggang}, journal={IEEE Transactions on Geoscience and Remote Sensing}, volume={61}, pages={1--16}, year={2023}, publisher={IEEE} }

@article{xiao2021kernel, title={A Kernel-Based Multi-Featured Rock Modeling and Detection Framework for a Mars Rover}, author={Xiao, Xueming and Yao, Meibao and Liu, Haiqiang and Wang, Jiake and Zhang, Lei and Fu, Yuegang}, journal={IEEE Transactions on Neural Networks and Learning Systems}, year={2021}, doi={10.1109/TNNLS.2021.3131206}, publisher={IEEE} }

@data{34a5-jq14-22, doi = {10.21227/34a5-jq14}, url = {https://dx.doi.org/10.21227/34a5-jq14}, author = {Xiao, Xueming and Yao, Meibao and Liu, Haiqiang}, publisher = {IEEE Dataport}, title = {MarsData-V2, a rock segmentation dataset of real Martian scenes}, year = {2022} } ```

References

[1] Xiao X, Yao M, Liu H, et al. A Kernel-Based Multi-Featured Rock Modeling and Detection Framework for a Mars Rover[J]. IEEE Transactions on Neural Networks and Learning Systems, 2021.

[2] Furln F, Rubio E, Sossa H, et al. Rock detection in a Mars-like environment using a CNN[C]//Mexican Conference on Pattern Recognition. Springer, Cham, 2019: 149-158.