Data entry ops · Production

NASA JPL uses Labelbox to build an AI-powered Martian frost map from multi-source imagery and thermal data

The problem

NASA JPL needed to construct an ML training dataset combining expert-annotated high-resolution imagery with thermal data to power a machine learning model for mapping frost on the Martian surface — a challenging task because visible indicators of frost vary widely and in subtle ways.

Workflow diagram · grounded in source
1
Multi-source data collection
trigger
“researchers collected, organized, and labeled high-resolution image data from multiple sources including the Mars Reconnaissance Orbiter. Lower-resolution image data would also be generated from various other cameras, and thermal data co…”
2
Image tiling and preprocessing
integration
“Because these images were originally so large, they were broken down into smaller squares and then randomized and labeled”
3
Three-labeler annotation via Labelbox
human_review
“the team had each image labeled by three labelers via Labelbox. Each labeler had to draw a polygon identifying frost, describe why they believe frost is present in the image, and rate their own confidence in the accuracy of their label”
4
Low-consensus group review
human_review
“The team then reviewed the images with low consensus scores as a group to improve consistency and identify specific causes of low confidence”
5
Iterative living labeling guide
feedback_loop
“the continuous creation of a labeling guide. As the team iterated on their labels, this guide was also updated with newly discovered edge cases and examples, so that the next set of labelers had reliable and accurate reference material f…”
6
ML model training on image tiles
ai_action
“They broke the images further down into 300px X 300px tiles, which were then used to train the initial model”
7
Performance testing and targeted curation
validation
“They then tested the model's performance on a separate dataset and measured areas of low confidence and performance. When specific terrains or formations that caused poor performance were identified, the team curated a new training datas…”
8
Thermal data integration
integration
“combining their labeled image data with thermal data collected by the Mars Climate Sounder. To do this, the team paired data points by metadata such as the time and location that the data was collected”
9
Dual-dataset confidence cross-referencing
validation
“you can take that model confidence and then cross it with the confidence that you have that frost exists based on a second dataset”
10
Dataset publication
output
“The frost map dataset was created over many iterations via Labelbox, and will soon be published to the broader scientific community, where other scientists and experts can ask questions, add context, and refine the data even more”
Reported outcome

The frost map dataset was created over many iterations via Labelbox and will be published to the broader scientific community for further refinement.

Reported metrics
Frost map dataset creationcreated over many iterations
Frost map release cadencereleased on a monthly basis
Reported stack
Labelbox AnnotateLabelbox
Source
https://labelbox.com/customers/nasa-jpl-martian-frost-map-customer-story/
Read source ↗

Frequently asked questions

What did this team achieve with this AI workflow?

The frost map dataset was created over many iterations via Labelbox and will be published to the broader scientific community for further refinement.

What tools did this team use?

Labelbox Annotate, Labelbox.

What results were reported?

Frost map dataset creation: created over many iterations; Frost map release cadence: released on a monthly basis (source-reported, not independently verified).

How is this data entry ops AI workflow structured?

Multi-source data collection → Image tiling and preprocessing → Three-labeler annotation via Labelbox → Low-consensus group review → Iterative living labeling guide → ML model training on image tiles → Performance testing and targeted curation → Thermal data integration → Dual-dataset confidence cross-referencing → Dataset publication.

Related data entry ops cases