Self-Supervised Learning to Guide Scientifically Relevant Categorization of Martian Terrain Images

@article{Panambur2022SelfSupervisedLT,
  title={Self-Supervised Learning to Guide Scientifically Relevant Categorization of Martian Terrain Images},
  author={Tejas Panambur and Deep Chakraborty and Melissa Meyer and Ralph E. Milliken and Erik G. Learned-Miller and Mario Parente},
  journal={2022 IEEE/CVF Conference on Computer Vision and Pattern Recognition Workshops (CVPRW)},
  year={2022},
  pages={1321-1331}
}
Automatic terrain recognition in Mars rover images is an important problem not just for navigation, but for scientists interested in studying rock types, and by extension, conditions of the ancient Martian paleoclimate and habit-ability. Existing approaches to label Martian terrain either involve the use of non-expert annotators producing taxonomies of limited granularity (e.g. soil, sand, bedrock, float rock, etc.), or rely on generic class discovery approaches that tend to produce perceptual… 

Figures and Tables from this paper

Mixed-domain Training Improves Multi-Mission Terrain Segmentation

This research proposes a semi-supervised learning approach that leverages unsupervised contrastive pretraining of a backbone for a multi-mission semantic segmentation for Martian surfaces that will expand upon the current Martian segmentation capabilities by being able to deploy across different Martian rover missions for terrain navigation.

Improving Contrastive Learning on Visually Homogeneous Mars Rover Images

Two approaches are proposed: an unsupervised deep clustering step on the Mars datasets, which identifies clusters of images containing similar semantic content and corrects false negative errors during training, and a simple approach which mixes data from different domains to increase visual diversity of the total training dataset.

Semi-Supervised Representative Region Texture Extraction of Fa\c{c}ade

The concept of Representative region texture is introduced, which can be used in the above modeling approach by tiling the representative texture around the façade region, and a semisupervised way to do representative region texture extraction from a façades is proposed.

References

SHOWING 1-10 OF 69 REFERENCES

Automatic Annotation of Planetary Surfaces With Geomorphic Labels

A methodology for automatic geomorphic mapping of planetary surfaces that incorporates machine-learning techniques is presented and can be adopted to generate geomorphic maps of sites on Earth.

DoMars16k: A Diverse Dataset for Weakly Supervised Geomorphologic Analysis on Mars

This work presents a method to automate a specific type of planetary mapping, geomorphic mapping, taking machine learning as a basis, and introduces a novel dataset, termed DoMars16k, which contains 16,150 samples of fifteen different landforms commonly found on the Martian surface.

GMSRI: A Texture-Based Martian Surface Rock Image Dataset

This paper introduces a new dataset called “GMSRI” that is a mixture of real Mars images and synthetic counterparts which are generated by GAN, and shows that GMSRI is much larger in scale and diversity than the current same kinds of datasets.

SPOC: Deep Learning-based Terrain Classification for Mars Rover Missions

SPOC has a promising potential for a wider range of future applications, such as the automated discovery of scientifically important terrain features on existing Mars orbital imagery, as well as traversability analysis for future surface missions to small bodies and icy worlds.

AI4MARS: A Dataset for Terrain-Aware Autonomous Driving on Mars

  • R. M. SwanDeegan J. Atha M. Ono
  • Computer Science
    2021 IEEE/CVF Conference on Computer Vision and Pattern Recognition Workshops (CVPRW)
  • 2021
This work created the first large-scale dataset, AI4Mars, for training and validating terrain classification models for Mars, consisting of ~326K semantic segmentation full image labels on 35K images from Curiosity, Opportunity, and Spirit rovers, collected through crowdsourcing.

LabelMars: Creating an extremely large Martian image dataset through machine learning

An automated terrain labelling and classification system based on state of the art machine/deep learning which enables keyword based search is developed and achieved 5000 annotated images from the Spirit, Opportunity and Rover navigation camera data bases.

Classification scheme for sedimentary and igneous rocks in Gale crater, Mars

Deep Texture Manifold for Ground Terrain Recognition

A parametric distribution in feature space in a fully supervised manner is learned, which gives the distance relationship among classes and provides a means to implicitly represent ambiguous class boundaries.

Classification of Martian Terrains via Deep Clustering of Mastcam Images

  • M. ParenteTejas Panambur
  • Computer Science
    IGARSS 2020 - 2020 IEEE International Geoscience and Remote Sensing Symposium
  • 2020
A clustering method that jointly learns the parameters of a neural network and the cluster assignments of the resulting features to the unsupervised training of image patches from data acquired by the mast cameras on the MSL Curiosity rover performs reasonably according to visual inspection of the cluster quality and simple clustering performance measures.
...