Skip to content

Methods for novelty detection in Mastcam multispectral images of the Mars surface


Notifications You must be signed in to change notification settings


Repository files navigation

Novelty Detection in Multispectral Mastcam Images


Science teams for rover-based planetary exploration missions like the Mars Science Laboratory Curiosity rover have limited time for analyzing new data before making decisions about follow-up observations. There is a need for systems that can rapidly and intelligently extract information from planetary instrument datasets and focus attention on the most promising or novel observations. Several novelty detection methods have been explored in prior work for three-channel color images and non-image datasets, but few have considered multispectral or hyperspectral image datasets for the purpose of scientific discovery. The Mastcam instrument is a multispectral imaging system that acquires images of the Mars surface from the mast of the Mars Curiosity rover in visible to near infrared wavelengths for the purpose of scientific study. We performed a study to compare the performance of four novelty detection methods---Reed Xiaoli (RX) detectors, principal component analysis (PCA), autoencoders, and generative adversarial networks (GANs)---to identify novel geology in Mastcam multispectral images.

This repository exists to assist in experimenting with the novelty detecton codebase developed for this study through Jupyter notebooks in a Jupyter Lab. There is a Jupyter notebook for each of the four methods that demonstrates how to use each method for novelty detection and evaluate performance for the Mastcam dataset. To make dependency managing easy, everything is setup using Docker.

Citation for this work: Kerner, H. R., Wagstaff, K. L., Bue, B. D., Wellington, D. F., Jacob, S., Horton, P., Bell III, J. F., Kwan, C., Ben Amor, H. (2020). Comparison of novelty detection methods for rover-based multispectral images. Under review.


Getting Started


To setup this you will need docker and docker-compose which can be found here! All other dependencies will be installed through the Dockerfile. Be sure to have the docker daemon running!

Pre-Docker Setup

Make a new directory for this system, enter it, and clone this repo

mkdir novelty_det
cd novelty_det
git clone (repo_url)

Make a new folder for data, enter the cloned repository and run to download the datasets

mkdir mcam_data
cd (repo_name)
python ../mcam_data

Docker Setup

Edit the .env file specify which directorty contains the data

# Provide data sets

Optional: Use the script to create a new Jupyter Lab password. You will need this to login once the service is deployed. Edit the ACCESS_TOKEN variable with the new SHA key. By default, the password is asdf.

Launching Docker

To launch docker, start from the repository's directory and start up the server

docker-compose up

Now you can open https://localhost:8888 to access the Jupyter Lab! If prompted for a password use asdf of the password you set in the .env file.

In the Jupyter lab instance, this codebase will be mounted in /home/jovyan/work and the data will be in /home/jovyan/data.

To bring the server down simply use

docker-compose down


Methods for novelty detection in Mastcam multispectral images of the Mars surface







No releases published


No packages published